* feat(web): two-step URL elicitation completion (#1415)
Replace accept-on-open with an explicit completion step for URL
elicitations. Opening the URL now only transitions the panel to its
waiting state; the elicitation resolves with `accept` only when the user
clicks the revealed "I've completed it" action, or with `cancel`.
- ElicitationUrlPanel: add `onComplete`, reveal the completion button and
relabel open→"Reopen in Browser" while `isWaiting`.
- PendingClientRequestModal: wire local `isWaiting`; open no longer sends
a response. Removed the accept-on-open comment that pointed here.
- Tests cover open→waiting→complete and open→cancel, and assert opening
alone does not resolve the elicitation.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(web): handle URL-elicitation-required (-32042) error path
Implements the spec's "URL mode with elicitation required error" flow
(2025-11-25). When a tools/call returns a `-32042`
UrlElicitationRequired error, the client now surfaces the required URL
elicitation(s) and retries the call once completed, instead of showing a
bare error.
Core (InspectorClient.callTool):
- Detect `-32042` (typed UrlElicitationRequiredError or generic McpError)
via the new `getUrlElicitationsFromError` helper.
- Surface each carried elicitation in order through the existing pending-
elicitation queue / modal (reusing the two-step "Open → I've completed
it" UI), then re-issue the original call once all are accepted. A
decline/cancel aborts with a clear message; a bounded retry count guards
a server that keeps returning the error.
- `ElicitationCreateMessage.completeIfPending()` lets the optional
`notifications/elicitation/complete` auto-advance/auto-accept an open
URL elicitation; the completion-notification handler now resolves
(was: only removed) the pending entry.
- Extracted `attemptToolCall` / `dispatchFailedToolCall` from the old
callTool body to support the retry loop without duplicating bookkeeping.
Web (App):
- A non-spec `-32042` with no elicitations (nothing to open) shows a
dedicated yellow toast linking to a new UrlElicitationErrorModal with
the raw error body, mirroring the output-schema-validation toast.
Tests: helper unit tests, completeIfPending unit tests, callTool
error-path retry/cancel/no-list/ordinary-error tests (injected fake
client), the new modal (test + stories), and an App toast test.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(web): abort URL-elicitation error path on a repeated URL
Guard the -32042 retry loop: if the server's retry error re-requests a
URL the user already completed earlier in the same call, completing it
again can't make progress, so callTool aborts with a new typed
UrlElicitationLoopError instead of re-prompting (which would loop until
the retry cap). The web layer detects it and shows a "URL elicitation
loop" toast naming the repeated URL rather than a generic failure.
Tests: core abort-on-repeat test (URL presented once, two attempts, then
loop error) and an App loop-toast test.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(web): let Copy-URL reach the URL-elicitation completion step
Address review of #1436: the "I've completed it" action was gated behind
clicking "Open in Browser" (only that set isWaiting), so a user who used
Copy URL to finish the flow externally could only Cancel. Copying the URL
now also moves the panel into the waiting state, exposing the completion
step. Also trims the now-redundant second sentence of the waiting hint.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* refactor(core): address review of error-path URL elicitation
Review follow-ups on #1436 (all non-blocking):
- Remove the leftover bare block scope in attemptToolCall (extraction
artifact); body de-indented one level.
- Log a warning when callTool gives up after the URL-elicitation retry
cap is hit, so a server that keeps demanding new elicitations is
diagnosable rather than looking like an ordinary failure.
- Assert the dispatch-once invariant on the loop-abort and cancel-abort
paths (exactly one failed toolCallResultChange event).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* style(core): fix attemptToolCall body indentation
Run prettier over the extracted attemptToolCall method body, which was left
over-indented by 2 spaces after the bare-block removal in 0754b43d. core/ is
outside clients/web's format:check scope, so npm run validate did not catch it.
Whitespace-only; the pre-existing line 658 wrap (also outside the format scope)
is left as-is, out of scope for this PR.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(core): settle pending elicitations on disconnect to avoid callTool hang
Address review of #1436: an error-path `awaitUrlElicitation` promise only
settled via respond, but `disconnect()` cleared `pendingElicitations`
wholesale — so a disconnect mid-elicitation (server crash, manual
disconnect, transport close) left `callTool` (and the UI's in-flight
toolCallState) hanging forever.
- Add `ElicitationCreateMessage.cancel()`: settles a pending elicitation
as `{ action: "cancel" }` without calling `onRemove` (the queue is
cleared by the caller, so it must not splice mid-iteration).
- `disconnect()` now cancels each pending elicitation before clearing the
array, and dispatches `pendingElicitationsChange` (it already did so for
samples) so the modal closes on a transport-close teardown too.
Tests: disconnect-mid-elicitation settles callTool (no hang); cancel()
unit tests (resolves-without-remove, no-op once responded).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(web): add per-server Roots config + client roots capability (#1425)
Add MCP roots support to the v2 web Inspector as a per-server config
section in the server settings form (no separate Roots tab/screen).
- Persistence: add `roots: Root[]` to `InspectorServerSettings` and
`StoredMCPServer`; thread through serverList converters (lift on read,
drop empty-uri rows / blank names on write, omit when empty) and the
remote server's read-side drop guard + `validateSettings`.
- Capability: `setupClientForServer` always advertises the roots
capability (`roots: cleanRoots(savedSettings?.roots ?? [])`) so the
server can issue `roots/list` and receive `roots/list_changed`.
- Live apply: edits to a connected server's roots are pushed via
`inspectorClient.setRoots(...)` once, on settings-dialog close — diffed
against the client's current roots — rather than per keystroke, to
avoid flooding the server with `roots/list_changed` / `roots/list`.
- Form: new Roots accordion section with a required URI and an optional
name per row (unlike v1, which was URI-only), + Add Root, and remove.
Tests cover the persistence round-trip (serverList + servers-route),
the form/modal section, and reuse the existing InspectorClient
roots/`setRoots` integration coverage.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* refactor(roots): address review — dedup guards/cleanRoots, preserve extra fields, test live-apply
Responding to the PR #1431 code review:
- Dedup `isRootArray`: drop the copy inside `validateSettings` and reuse the
module-scope guard the read path uses, so the two can't drift
(core/mcp/remote/node/server.ts).
- Dedup the clean-roots normalizer: hoist a single exported `cleanRoots`
into core/mcp/serverList.ts and reuse it from both
`inspectorSettingsToStoredFields` and App.tsx (drops the local copy).
- Preserve non-form root fields (e.g. `_meta`): `cleanRoots` now spreads the
carried-through fields and only normalizes uri/name, and the modal's
`handleRootChange` spreads the existing root on edit — so a hand-edited
root's `_meta` survives load + persist + edit instead of being silently
dropped.
- Test the App live-apply-on-close path: new App.test.tsx cases assert
setRoots is called once with cleaned roots when changed on the active
server, and not called when the cleaned roots already match what the
client advertises. Adds a cleanRoots unit test (blank drop + _meta
preserve).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Wires the ported Tasks screen + core task subsystem through App.tsx: sticky Clear Completed, cancel/refresh error toasts, core-side per-task progress correlation, Run-as-task on the Tools screen (gated on server support), per-server Task TTL, and live task-status toasts.
* feat(web): tool progress toasts + persist tool selection/result across navigation (#1414)
Surface incoming notifications/progress as toasts so the user can watch a
long-running tool's progress without leaving the tool view — the v2 replacement
for v1's always-visible "Server Notifications" shelf. Toasts are keyed by
progress stream and updated (not stacked) per tick so a chatty server updates
one toast rather than flooding the corner. Progress notifications still land in
the History tab via the existing transport-level tracking.
Companion change: lift the selected tool and its form values into App.tsx
alongside the already-lifted toolCallState, make ToolsScreen controlled, and
drop the clear-on-unmount effect. Leaving the Tools tab and returning now
preserves selection, inputs, and result. All three reset together on
disconnect via resetSessionScopedUiState, preserving the #1368/#1394 behavior.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(web): persist per-screen selection/search/filter across navigation (#1417)
Lift every screen's selection, form, search, and filter state out of the
screen components into session-scoped state owned by App, passed down as
controlled props. The screens unmount on tab switch, so this is what keeps
the user's context in place when they navigate away and back within a live
session. State is cleared only on disconnect (resetSessionScopedUiState) or
an explicit user action — never on plain navigation.
Lifted per screen:
- Tools: sidebar search (selection/form/result were lifted in #1414)
- Prompts: selected prompt, argument values, submitted marker, search
- Resources: selected resource/template URI, originating-template marker,
search, accordion open-sections
- Apps: selected app, form values, search (running/maximized stay local —
they're tied to the live iframe/bridge that is torn down on unmount)
- Tasks: search + status filter
- Logs: filter text + visible-level set
- History: search + method filter
- Network: filter text + visible-category set
The four sidebar *Controls (Tool/Prompt/App/Resource) become controlled for
search; ResourceControls also controls accordion open-sections. Level/category
default maps move to logLevels.ts / fetchCategories.ts so the screen files
keep a single component export.
Pagination already persists (it lives in the App-level Paged*State managers),
so no work was needed there.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(web): persist list scroll position across navigation + reset tests (#1417)
Add useScrollMemory: a module-keyed hook that remembers a Mantine ScrollArea's
offset across unmount/remount (the screens unmount on tab switch). Scroll lives
in a module-scope map rather than threaded App state — it's ephemeral DOM state,
and threading ~16 more props would balloon the surface for no benefit. App calls
clearScrollMemory() in resetSessionScopedUiState so offsets clear on disconnect
alongside the lifted selection/filter state.
Wired into the four main list panels (Logs, History, Network, Tasks) and the
Tools/Prompts/Apps sidebar lists. Pagination already persists via the App-level
Paged*State managers.
Tests: useScrollMemory unit test (save/restore, key isolation, clear), and the
App disconnect test now also exercises representative lifted UI state (tool
search, prompt selection, log filter) clearing on disconnect.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* refactor(web): collapse per-screen lifted state into one {screen}Ui object
Replace the ~46 individual selection/search/filter props threaded App →
InspectorView → screen with one `{screen}Ui` object + one `on{Screen}UiChange`
setter per screen (8 pairs). Each screen reads its fields from `ui` and updates
via `onUiChange({ ...ui, field })`; the in-flight result panels
(toolCallState / getPromptState / readResourceState) stay separate since
they're written by App's async action handlers, not the screens.
The `*UiState` interfaces live with their screens; the empty-state defaults are
gathered in screenUiState.ts so the screen files keep a single component export
(react-refresh), mirroring logLevels.ts / fetchCategories.ts.
Trims InspectorView's prop surface substantially and keeps the persistence
behavior identical. Tests/stories collapse their stateful wrappers to a single
ui object; added Tools/Tasks/History interaction tests to keep per-file
coverage above the gate.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(web): dismiss visible progress toasts on client teardown (#1414 review)
The progressNotification effect cleanup cleared its id-tracking set but left
on-screen "Tool progress" toasts to auto-close up to 5s later — they could
linger into the next session, and the lingering toast's onClose could delete an
id from the new session's set and trigger a duplicate-id re-show. Hide each
tracked toast in cleanup so both gaps close. Addresses @claude review note #2.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(web): clear the tool result when a different tool is selected
The Tools result panel renders `toolCallState` regardless of which tool is
selected, so selecting a new tool used to leave the previous tool's result
showing beneath it. Route tools UI changes through an App handler that drops
`toolCallState` whenever `selectedToolName` changes; search/form edits keep the
selection (and the result) intact. Tests cover both the clear-on-switch and
keep-on-search paths.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The Hono dev-proxy middleware pumped streaming /api responses to the Node
ServerResponse without handling client disconnect. When the browser dropped a
stream mid-flight (SSE close, navigation, HMR reload) `res` was destroyed and
the next write failed with ERR_STREAM_DESTROYED, which the pump logged as an
error. It also had no `error` listener (a failed write could surface as an
uncaught exception) and no `close` listener (it kept reading upstream after the
client left).
Make the pump disconnect-aware:
- Listen for res `close`/`error`; either cancels the upstream reader (resolving
the pending read so the loop exits) and flips a `clientGone` flag.
- Treat ERR_STREAM_DESTROYED / EPIPE / ECONNRESET as a normal end — not logged.
- Only log genuinely unexpected stream errors.
- Await each write so a slow client applies backpressure instead of the proxy
buffering the whole stream.
Dev-only plugin (excluded from the coverage gate — it needs a real Vite server
to exercise). Verified via npm run validate.
Closes#1418
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(web): render pending sampling/elicitation requests (#1407)
Server-initiated sampling and elicitation requests never reached the v2
web UI: the InspectorClient queued them and returned a Promise that only
resolved on a user response, but nothing subscribed to the pending-change
events or rendered the (already-built) request components. A tool that
triggered one spun the Execute button forever.
Wire the queues to the UI:
- Add usePendingClientRequests() (core/react) subscribing to
pendingSamplesChange / pendingElicitationsChange and returning the live
arrays. Expose getPendingSamples() / getPendingElicitations() on
InspectorClientProtocol (and the test fake) so hooks can read them.
- Add PendingClientRequestModal, an App-level modal that surfaces one
queued request at a time (sampling-first) via the existing
SamplingRequestPanel / ElicitationFormPanel / ElicitationUrlPanel. It is
non-dismissable so the blocked call only resolves via an explicit action.
- Wire it in App.tsx: responding/rejecting resolves the client's handler
Promise, which unblocks the originating call and clears the spinner.
Also fold in two issues surfaced while verifying:
- SchemaForm rendered an `array` + `items.enum` (multi-select enum) as a
JSON textarea because only `items.anyOf` had a branch; add a MultiSelect
branch with enumNames label support.
- ToolsScreen's result lives in App and outlived the screen's local
selection (reset on tab-switch unmount), leaving a result with no
selected tool. Clear the result on unmount for a clean slate on return.
Prompts/Resources gate their preview on selection, so they are unaffected.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(elicitation): support explicit decline; honest queue label; review polish
Address @claude review on #1411:
- Add an explicit "Decline" action to ElicitationFormPanel so the spec's
decline (refusal) is distinguishable from cancel (dismissal); the JSDoc
claimed accept/decline/cancel but only accept/cancel were wired. Wire the
modal body to emit { action: "decline" }.
- Replace the misleading "1 of N" queue label (the leading 1 never changed
since the modal always shows the head) with an "N pending" hint, hidden when
nothing else is queued.
- Make the dead useMemo honest: derive the head request inside the memo from
the source arrays so it actually caches (the inline activeElicitation had a
fresh identity every render).
- Document the URL elicitation accept-on-open semantic with a comment pointing
at the follow-up (#1415).
- Add an App-level round-trip test driving pendingSamplesChange end-to-end
(modal opens, respond() fires, empty event closes it).
Follow-ups filed: #1415 (URL two-step completion flow), #1416 (form required-
field submit gate + double-submit guard).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(elicitation): include untouched schema defaults in submitted content
The schema form displays a field's default via resolveValue(), but onChange
only writes a field into `values` once the user edits it. So an elicitation
submitted with default-only fields left untouched dropped them entirely —
v2 sent just the manually-edited fields while v1 sent defaults too (reported
with the everything server: firstLine/integer/number/enum defaults missing).
Add collectSchemaDefaults() to jsonUtils (recurses nested objects, omits
fields with no default) and seed ElicitationFormModalBody's values state with
it via a lazy initializer. The modal body is keyed by request id, so it
re-seeds per request; user edits still override seeded defaults via the
existing onChange merge.
Tests: collectSchemaDefaults unit tests (field types, nested, empty) in
jsonUtils.test; a modal test asserting an untouched-default form still submits
its defaults in content.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(forms): seed schema defaults for tool-call and app-launch forms too
Extend the elicitation defaults fix (collectSchemaDefaults) to the other two
SchemaForm consumers that own form state and submit it: ToolsScreen (tool
execute args) and AppsScreen (app launch args). Both reset formValues to {} on
selection, so default-only fields the user never edited were dropped from the
submitted args — same root cause as the elicitation modal.
Seed formValues from the selected tool/app inputSchema defaults on select.
User edits still override via the existing onFormChange merge.
(InlineElicitationRequest has no consumers wired up, so it's left as-is.)
Tests: ToolsScreen sends a tool's default on Execute without editing;
AppsScreen sends an app's default on Open App without editing.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(elicitation): gate form submit on required fields; guard double-submit (#1416)
Fold the #1416 hardening into this PR:
- Gate the elicitation form's Submit button on required fields. Promote the
hasMissingRequiredFields helper out of AppDetailPanel into jsonUtils (shared)
and use it in ElicitationFormPanel so Submit is disabled until every required
field has a value.
- Guard against double-submit. Add an optional `busy` prop to the sampling,
elicitation-form, and elicitation-url panels that disables their responding
actions. PendingClientRequestModal's bodies use a new ref-guarded
useRespondOnce() hook: the first action dispatches and locks the panel; later
clicks no-op. This stops a fast double-click from calling respond()/reject()
twice (which throws and produced an unhandled rejection). Copy URL stays
enabled — it doesn't resolve the request.
Tests: hasMissingRequiredFields unit tests; ElicitationFormPanel submit-gate
and busy-lock tests; a modal double-click-resolves-once test.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(tools): cap detail/result panels at viewport and scroll internally
The Tools screen's detail and result columns sized to their content and bled
past the bottom of the viewport on long output (e.g. the env dump or a long
elicitation result). Cap each column at the screen's available height and let
its panel scroll internally instead.
- ToolsScreen: drop `align: flex-start` and wrap each column in a stretched
`ContentPane` (Flex column, capped by `mah` = viewport − header − xl*2). The
pane's definite height is what lets the card's inner ScrollArea know how far
to shrink — a bare `mah` on the card does not (the card sizes to content, so
the scroll region renders full-height and overflow:hidden just clips it,
unreachable). Mirrors the proven Prompts/Resources preview panes.
- ToolResultPanel: pinned header + inner ScrollArea (flex 0 1 auto, mih 0).
- ToolDetailPanel: pinned title/annotations header, scrolling body
(description + form + progress), pinned Execute footer.
Audited the other screens: Prompts/Resources already use this pattern,
ServerList and the four list panels (Logging/Network/History/Tasks) already cap
via ScrollArea.Autosize. Only AppsScreen's app-input form had the same gap —
fixed it the same way: AppDetailPanel now scrolls internally and the card uses
variant="preview" so the running app's iframe and a long form both stay within
the viewport.
Verified in-browser against the everything server: the get-env result now
stops at the viewport bottom with margin and scrolls within the card
(clientHeight 502 vs scrollHeight 1931), header pinned.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* test(remote): use sync pino destinations in logging tests to fix teardown race
CI flaked with an uncaught `ENOENT: ... /tmp/inspector-log-unknown-level-*/app.log`
that failed an unrelated later test. The remote-logging integration tests create
`pino.destination(logFile)` loggers (async sonic-boom streams) and rmSync the temp
dir in finally/afterEach without closing the stream — a buffered background flush
then fires after the dir is gone and throws an uncaught error, which Vitest counts
as a run failure even though all tests pass.
Make those destinations `sync: true` so the write lands during the request instead
of on a later flush, removing the race. Test-only; unrelated to this PR's feature
work but it was blocking the PR's CI.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(storage): await store-file writes instead of polling (#1412)
OAuth persistence is Zustand `persist` → file-storage adapter →
writeStoreFile, which the persist middleware invokes fire-and-forget:
saveTokens updates the in-memory store synchronously while the file
write lags. Tests verified persistence by polling the state file with a
2s/50ms `waitForStateFile()` budget, which flaked under full-suite load
(the file was written, the poll just didn't observe it in time).
Make the write completion awaitable instead:
- Track in-flight writeStoreFile() promises per resolved path in
store-io and add flushStoreFileWrites(filePath?) to await them (also
useful for graceful shutdown). Writes to the same path are chained so a
flush awaits every queued write.
- Convert the OAuth E2E, node storage, and file-storage-adapter tests
from poll loops (waitForStateFile / vi.waitFor on existsSync /
setTimeout) to `await flushStoreFileWrites(path)` + a single read.
- Remove the now-unused waitForStateFile() test helper.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* chore(storage): doc load-bearing sync registration; rename stale debug env
Address @claude review on #1413:
- Document that writeStoreFile()'s synchronous pendingWrites registration is
load-bearing for flushStoreFileWrites(), in both the storage adapter's
setItem callsite and the pendingWrites docblock, so future refactors don't
introduce an await before registration and silently re-flake the tests.
- Rename the stale DEBUG_WAIT_FOR_STATE_FILE test env var to DEBUG_PERSIST_PATH
(waitForStateFile was removed in this PR).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* chore(web): silence benign config-load dep warnings in `npm run dev`
`vite dev` printed three UNRESOLVED_IMPORT warnings at startup (chokidar,
atomically, @napi-rs/keyring). They originate from Vite's config bundler
(Rolldown): the Hono plugin statically imports the node-only dev backend
(core/mcp/remote/node/server.ts), so loading vite.config.ts bundles it and
Rolldown walks its node-only deps. Those deps are correctly externalized — the
browser never loads them and Node resolves them at runtime — so the warnings are
pure noise.
They can't be filtered the usual ways: the static import is required (the
monorepo has no node_modules under core/, so the aliases must resolve the deps
at bundle time — ssrLoadModule / `--configLoader runner` both fail to resolve
them), and the warnings are written directly to the process streams during
config load, bypassing customLogger, logLevel, and rollupOptions.onwarn.
Fix: a tiny `--import` hook (server/quiet-config-warnings.mjs) that drops only
those specific benign warning writes, wired into the `dev` script. Scoped to
`npm run dev` — build, tests, and prod are untouched.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* chore(web): harden dev warning filter per review
- Tighten BENIGN_PAIRS to repo-relative path fragments (not basenames) so an
unrelated future UNRESOLVED_IMPORT warning mentioning the same dep + a
same-named file isn't silently dropped. Verified Rolldown prints the full
repo-relative path. (review #1)
- Add a note that the matcher tracks Rolldown's message format and degrades to a
safe no-op (warnings return) if that format changes. (review #2)
- Export isBenignWarning/BENIGN_PAIRS, guard the stream patching behind
!VITEST, add a .d.mts so the pure helper is importable from TS, and add a unit
test locking the signature pairs and the over-match guards. (review #3)
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(apps): render running MCP Apps end-to-end (#1370)
Wire the v2 Apps screen so clicking Open App actually runs an MCP App.
Previously App.tsx stubbed the three load-bearing pieces (sandbox URL,
bridge factory, action handlers), so the iframe loaded about:blank and
nothing crossed the bridge.
- Add useSandboxUrl hook (core/react) to read sandboxUrl from GET
/api/config and thread it down as sandboxPath; render an "MCP Apps
unavailable" state instead of a blank iframe when absent.
- Add createAppBridgeFactory: builds a host-side ext-apps AppBridge over
the active client's SDK client, reads the tool's UI resource on the
sandbox handshake and pushes it into the proxy, handles openLinks, and
connects a PostMessageTransport.
- AppRenderer: pass the tool to the bridge factory; buffer tool
input/result and release them once the view signals `initialized`
(input before result, latest-wins).
- Move the AppRenderer handle ref to App.tsx so onOpenApp issues
tools/call with the form values and delivers input + result to the
running app via the bridge; onCloseApp tears the bridge/iframe down.
- Stories: themed sandbox placeholder (visible in light/dark), an
echo-bridge running story, and an unavailable story. New unit tests
for the factory, the buffering, useSandboxUrl, and the empty state.
The listChanged indicator badge wiring is tracked separately in #1402.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(apps): make MCP Apps render in dev — single bridge + result passthrough
Two bugs kept apps (e.g. NightRider's MIDI Pattern Editor) stuck on an empty
shell even though the host fetched the resource and called the tool correctly.
1. StrictMode double-mounted the bridge. The dev-only effect setup→cleanup→
setup built TWO host bridges/transports, double-loaded the sandbox, and
raced the app's ui/initialize handshake (it timed out, so `initialized`
never fired and buffered input/result were never flushed). AppRenderer now
defers bridge disposal to a microtask and, when the effect re-runs
synchronously with identical inputs, cancels the disposal and reuses the
live bridge — so a single bridge survives the double-invoke. The pending
input/result buffer is no longer cleared on a transient cleanup.
2. The SDK rejected the server's result. callTool() validates structuredContent
against the tool's outputSchema and throws on a mismatch — but an MCP App is
a passthrough: the host must forward whatever the server returned to the
running view (which legacy hosts render fine). InspectorClient.callTool gains
a `skipOutputValidation` option that routes through client.request() (no
wrapper validation); onOpenApp uses it.
Verified end-to-end against a real app via Playwright: the editor now renders
its grid + pattern data identically to the v1 inspector.
Tests: AppRenderer StrictMode single-bridge/no-dispose case; a get_temp_extra
test-server fixture (returns an extra structuredContent property) with
integration tests covering both the throwing default path and the
skipOutputValidation passthrough.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(apps): warn (toast) when an app's result violates its outputSchema
The skipOutputValidation passthrough keeps schema-mismatched apps rendering,
but that silently hid a real spec problem from server developers. Now the
host still renders, but surfaces the mismatch.
- InspectorClient.callTool runs the SDK's own Ajv output validation on the
skipOutputValidation path NON-fatally and returns the message as
`outputValidationError` on the ToolCallInvocation (and the
toolCallResultChange event); the result is still delivered.
- App.tsx onOpenApp shows a yellow toast when outputValidationError is set,
explaining strict MCP clients may not render the app.
Tests: integration cases asserting outputValidationError is set for a
schema-violating result (get_temp_extra) and unset for a valid one (get_temp).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(apps): summarize output-schema warning in toast + details modal
The previous toast dumped the full (very long) validation error and
overflowed. Now the toast shows a one-line summary with a "View validation
details" link that opens a modal containing the full error in a read-only,
scrollable textarea.
- New OutputValidationModal group component (+ story + tests).
- App.tsx: toast message is a summary + link; clicking it opens the modal
populated with the tool name and the full outputValidationError.
Verified via Playwright against the NightRider MIDI editor: toast is concise,
link opens the modal with the scrollable details.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* refactor(apps): address PR review — theme stability, onError, validation polish
Responds to the automated PR review on #1403.
- Theme toggle no longer rebuilds a running app's bridge. createAppBridgeFactory
reads the resolved theme from the DOM (Mantine's data-mantine-color-scheme) at
bridge-build time instead of capturing isDark in the factory deps, so the
factory identity is stable across theme flips. (#1)
- Wire AppRenderer's onError through AppsScreen -> InspectorView -> App.tsx,
surfacing bridge/runtime failures (no client, late rejection, failed
tools/call) as a red toast instead of a silent blank iframe. (#2)
- Extract validateToolOutput into a pure, unit-tested helper covering the
matching / extra-property / missing-structuredContent / error / bad-schema
cases (the no-structuredContent branch can't be produced via the SDK server,
which rejects it before sending). (#4)
- useSandboxUrl aborts its in-flight fetch on unmount (cancelled flag) to avoid
setState after unmount. (#5)
- Schema-mismatch toast: autoClose:false (details modal is one click away). (note)
- AppRenderer: add a state-machine comment block documenting the bridge
lifecycle refs. (#6)
- inspectorClient: note why the `as CallToolResult` casts are safe. (note)
Tests: AppsScreen onError surface; validateToolOutput unit suite; factory theme
read from DOM. validate + test:storybook green.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* docs(apps): clarify theme attribute + useSandboxUrl fetch comments
Re-review follow-ups (non-blocking):
- createAppBridgeFactory: note that Mantine's data-mantine-color-scheme is only
ever light/dark (auto is resolved before paint), so no `auto` branch is needed
and matchMedia only covers an absent attribute.
- useSandboxUrl: correct the doc to say it fetches on mount AND re-fetches on
baseUrl/authToken change (idempotent), not "once on mount".
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(state): dispatch connect event after capabilities are populated
The list-state managers (tools/prompts/resources/templates) refresh on the
"connect" event and, since #1395, gate their list RPC on getCapabilities().
But connect() dispatched "connect" before fetchServerInfo() populated
this.capabilities, and refresh() runs synchronously up to its first await —
so the capability gate executed during the synchronous dispatch, read
undefined capabilities (reset to undefined on every disconnect), tripped,
and wiped every list to empty on connect. Tools, prompts, resources, and
resource templates all vanished against any server.
Move the "connect" dispatch to after the awaited fetchServerInfo() so
capabilities are in place before any manager refreshes. The unit tests
missed this because their FakeInspectorClient has capabilities populated
from construction; added an integration regression test wiring a real
ManagedToolsState on connect and asserting the tool list is non-empty
(fails without the fix).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* test(integration): harden connect-race regression test per review
Address the two non-blocking suggestions from the PR review:
- Replace the 100ms settle sleep with waitForEvent() on the managers'
change events, so the test is deterministic against slow CI instead of
racing a fixed timeout.
- Assert a second affected primitive (ManagedResourcesState alongside
ManagedToolsState). The fix is a single shared line, but covering two
distinct capabilities guards against a future change gating one
primitive differently.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Pull the four session-scoped UI resets (toolCallState / getPromptState /
readResourceState panels + optimistic currentLogLevel) out of the
disconnect listener into a colocated `resetSessionScopedUiState`
useCallback, so there's a single named place to extend as App.tsx accrues
more per-session state. The disconnect listener keeps its activeServerId +
modal-flag bookkeeping inline and calls the helper. Pure refactor — no
behavior change; the existing App.test.tsx disconnect test (asserts all
four pieces reset) is unchanged and still passes.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The four list-fetching state managers fired their list RPC against every
server on connect regardless of advertised capabilities, so a server that
doesn't implement a given primitive replied -32601 "Method not found",
surfacing in the console on every connect. Mirrors the `tasks`-capability
gate added for ManagedRequestorTasksState in #1349.
- managedToolsState → gate on capabilities.tools
- managedPromptsState → gate on capabilities.prompts
- managedResourcesState → gate on capabilities.resources
- managedResourceTemplatesState → gate on capabilities.resources
(the spec defines no separate resourceTemplates capability;
resources/templates/list is part of the resources surface)
When the capability is absent each manager sets an empty list, dispatches
its change event, and returns — the right semantics for "this server
doesn't support X."
Tests: existing flow tests now construct their FakeInspectorClient with
the relevant capability so the live list path is still exercised; added a
"refresh skips listX when capability absent" and a "connect against an
X-less server doesn't fire listX" test per manager. Updated the
useManagedX peers and resourceSubscriptionsState (which drives a
ManagedResourcesState refresh) to advertise the capability too.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(app): clear per-call panels and log level on disconnect (#1368)
Session-scoped UI state owned by App.tsx — the Tools / Prompts /
Resources result panels (`toolCallState` / `getPromptState` /
`readResourceState`) and the optimistic `currentLogLevel` — survived a
disconnect/reconnect cycle, so server B's screens showed server A's last
result. Extend the existing `disconnect` listener to reset all four back
to empty / "info", alongside the activeServerId + modal-flag resets it
already does. This covers all three disconnect paths the listener handles
(explicit toggle, header Disconnect button, mid-session transport
failure). `latencyMs` continues to reset via the connectionStatus effect.
Adds App.test.tsx covering the disconnect → tool-call-panel-cleared and
log-level-reset transition, mocking the InspectorClient (a fake
EventTarget), the per-server state managers, the core hooks, and
InspectorView.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* test(app): assert all four panels reset on disconnect (#1368)
Address PR review: the disconnect test asserted only toolCallState and
currentLogLevel. Extend the InspectorView double with prompt/resource
status spans and get-prompt/read-resource buttons so the test fills and
then asserts the reset of all four pieces of session-scoped state
(tool / prompt / resource panels + log level), fully matching the
"reset on disconnect" contract. Retitle the test accordingly.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(auth): inject MCP_INSPECTOR_API_TOKEN into served index.html (#1378)
Reloading the web client at the bare URL (no `?MCP_INSPECTOR_API_TOKEN=…`
query string) with empty sessionStorage made every `/api/*` request 401 —
the browser had no way to recover the backend's auth token.
Embed the token into `index.html` on every page load so the browser no
longer depends on the query string surviving navigation:
- New shared helper `clients/web/server/inject-auth-token.ts` embeds
`<script>window.__INSPECTOR_API_TOKEN__ = "…"</script>` (escaped against
`</script>` injection; no-op when auth is dangerously omitted).
- Dev: the Vite plugin injects via `transformIndexHtml`.
- Prod: the Hono server injects on the `/` route.
- `App.tsx` `getAuthToken()` now reads the injected global first, then the
query string, then sessionStorage (both fallbacks preserved).
- Shared global name lives in `INSPECTOR_API_TOKEN_GLOBAL`
(`core/mcp/remote/constants.ts`).
Tests: helper unit coverage + an integration test exercising the real
prod server's `/` → `/api/*` flow (injected token authenticates; missing
token 401s). AGENTS.md documents the token-recovery order.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(auth): wire OAuth authorization-code flow into App.tsx (#1379)
OAuth-protected MCP servers could not be connected to from the v2 web
client: the core OAuth pipeline exists, but App.tsx never invoked it, so a
connect attempt 401'd and surfaced "Remote send failed (401): … Missing
Authorization header" as a toast.
Wire the two missing entry points (all core primitives already in place):
- Auto-trigger on 401: in onToggleConnection's catch, detect an upstream
401 (isUnauthorizedError) and call client.authenticate(), which runs
discovery + DCR (backend-proxied) and redirects the page to the auth
server via BrowserNavigation. The initiating server id is persisted to
sessionStorage first, since the OAuth `state` carries only mode+authId
and the full-page redirect wipes React state.
- /oauth/callback handler: a mount effect that, once `servers` hydrate,
parses the callback params, recovers the pending server, rebuilds its
InspectorClient, runs completeOAuthFlow(code) (PKCE verifier + DCR client
info survive in BrowserOAuthStorage), replaceState("/") so a reload can't
replay the single-use code, then connect(). An `error=` callback toasts
instead of retrying.
connect() already attaches the OAuth provider to the transport
(inspectorClient.ts), so once tokens land in BrowserOAuthStorage the
outbound request carries the bearer token.
Extracted the pure pieces (constants + isUnauthorizedError) to
src/utils/oauthFlow.ts with unit tests. Verified end-to-end in a real
browser against the MCP SDK demo OAuth server: Connect -> redirect ->
auto-approve -> callback -> Connected, with the access token shown in the
Connection Info modal (#1377).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(auth): persist OAuth pre-redirect Network log across the redirect (#1384)
After #1379, the Network tab showed only the post-redirect auth HTTP
(discovery re-run + POST /token); the pre-redirect discovery and the DCR
POST /register that run during authenticate() were lost when the page
navigated to the auth server.
Root causes:
1. Ordering — BrowserNavigation set `window.location.href` before the
client's `saveSession` event fired (OAuthManager calls onBeforeOAuthRedirect
*after* auth() already navigated), so the save raced the unload and was
dropped. Fix: BrowserNavigation now runs a synchronous `beforeNavigate`
hook immediately before assigning location.href; App wires it through
createWebEnvironment to flush the active fetch log to RemoteInspectorClient
Storage (keyed by the authId parsed from the auth URL) via a keepalive POST
that outlives the unload.
2. Illegal invocation — RemoteInspectorClientStorage defaulted to
`this.fetchFn = globalThis.fetch` and called `this.fetchFn(...)`, which
re-binds `this` and makes native fetch throw "Illegal invocation"
(swallowed by the catch). This silently broke *all* session save/load.
Fix: default to a wrapper that preserves the global receiver.
3. Restore race — hydrateFetchRequests replaced the list, so a load that
resolved after the resuming connect appended live entries would clobber
them. Fix: merge restored (older) entries ahead of live ones, dedupe by id.
saveSession also now uses keepalive: true.
Verified end-to-end against the MCP SDK demo OAuth server: the connected
page's Network tab shows the full handshake — pre-redirect discovery + DCR
/register plus post-redirect discovery + /token as `auth`, alongside
`transport`. Added unit tests for the beforeNavigate ordering and the
hydrate merge/dedupe.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(network): show auth response bodies with masked secrets + reveal toggle (#1386)
Auth-category Network entries showed request body/headers/status but never
the response body (rendered "(empty)"), because buildEffectiveAuthFetch
deliberately skipped capturing it to avoid surfacing access_token /
refresh_token. That hid the most useful thing to inspect when debugging
OAuth — the token exchange.
Capture auth response bodies, but mask sensitive OAuth fields by default
behind a click-to-reveal toggle so they aren't exposed at a glance during
a screen-share:
- inspectorClient: wire updateResponseBody on the auth fetcher.
- src/utils/maskSecrets.ts: maskSecretsInBody() masks access_token,
refresh_token, id_token, client_secret (case-insensitive, nested) in JSON
bodies; reports whether anything was masked. Non-JSON / secret-free bodies
pass through untouched.
- NetworkEntry BodyPreview: when a body has masked fields, render it masked
with a Reveal/Hide toggle (copy honors the shown view). Masking is a UI
concern; the raw entry is unchanged so reveal shows the real value.
access_token / refresh_token live in the post-redirect /token response,
which is never written to the session-restore files (#1384); only
pre-redirect bodies (public discovery, DCR /register) persist, so no bearer
token hits disk.
Verified end-to-end: the /token response shows masked by default
(access_token: "••••••••"), Reveal exposes the raw value, and discovery /
the public DCR /register response (no secret) render with no toggle. Added
util + component unit tests and a story play function.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(auth): inject token into prod SPA fallback + prime sessionStorage + no-store (#1378)
Addresses code-review feedback on the token-injection PR:
- Prod `server.ts` SPA deep-link fallback (e.g. `/oauth/callback`) previously
served the raw index.html off disk via serveStatic, bypassing injection — so
a bookmark/reload at a non-`/` route with empty sessionStorage would 401.
Route the SPA fallback through the same `serveIndexHtml` (inject) helper;
real static assets (paths with a dot) still serve verbatim. Dev already
injected on every HTML serve via Vite `transformIndexHtml`.
- `getAuthToken()` now persists the injected `window.__INSPECTOR_API_TOKEN__`
to sessionStorage (not just the URL-param branch), priming the backstop for
any later navigation that loses the global.
- Injected HTML responses now send `Cache-Control: no-store`, so a page
carrying a token isn't cached and served stale after a restart regenerates
the token.
Integration tests added: SPA fallback (`/oauth/callback`) carries the token,
`Cache-Control: no-store` on injected HTML, real assets served verbatim, and
unknown `/api` routes 404 rather than falling through to the HTML shell.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(auth): address #1383 review — split OAuth/connect toasts, tighten 401 match (#1379)
Code-review feedback on the OAuth-wiring PR:
- Callback effect: split completeOAuthFlow vs connect() into separate
try/catch blocks. A token-exchange failure now reads "OAuth token exchange
failed … Please try connecting again." (the single-use code is spent and the
URL was cleared, so a reload can't retry); a post-OAuth connect failure reads
"Failed to connect" since OAuth actually succeeded and re-clicking Connect
reuses the persisted tokens.
- isUnauthorizedError: anchor the message fallback on the transport's
`failed …(401)` wording instead of a bare `(401)`, so an unrelated `(401)`
spliced into an error message can't trip the OAuth flow. Added a test.
- Documented that clearing the pending id + URL before the server lookup is
intentional (deleted/renamed server mid-flow → require a fresh Connect).
Also merges the squash-merged #1382 base from v2/main.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* docs(auth): address #1385 review — clarify double-save + keepalive cap + test fetch default (#1384)
Code-review feedback on the OAuth Network-log persistence PR:
- Documented the double-save: `FetchRequestLogState`'s `saveSession` listener
is the backstop; `BrowserNavigation`'s `beforeNavigate` hook is the primary
flush for the redirect case. Notes the listener may lose the navigation race
and is harmless when it duplicates (last-writer-wins, identical payload).
- Reworded the keepalive comment in `RemoteInspectorClientStorage.saveSession`:
the 64KB cap is general (the method is also reachable from the listener with
the full session log), so a long session could exceed it and drop silently —
acceptable since the persisted log is best-effort, not load-bearing.
- Added a regression test that constructs `RemoteInspectorClientStorage`
without a `fetchFn`, stubs `globalThis.fetch`, and asserts the default
wrapper calls it (locks in the "Illegal invocation" fix, which callers
otherwise swallow).
Optional items (logger.warn on swallowed save errors; setupClientForServer dep
churn) acknowledged on the PR, not changed.
Also merges the squash-merged #1383 base from v2/main.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(network): address #1387 review — form-encoded masking, cleaner flag, a11y (#1386)
Code-review feedback on the auth-response-body masking PR:
- Extended masking to form-urlencoded bodies (finding #1): the token *request*
is `application/x-www-form-urlencoded` and carries `code` / `code_verifier` /
`client_secret` / `refresh_token`. `maskSecretsInBody` now masks those in
form bodies too (preserving formatting; placeholder not percent-encoded).
`code`/`code_verifier`/`client_assertion` are form-only sensitive keys —
deliberately NOT masked in JSON, where `code` is usually an error/status code.
- Replaced the double-stringify `hasSecrets` heuristic with an explicit
masked-flag propagated out of `maskNode` (finding #2) — robust if the
transform ever grows non-identity behavior, and one fewer serialization.
- Reset reveal state on body change via `key={body}` remount instead of a
setState-in-effect (finding #3; avoids the cascading-render lint rule).
- a11y (finding #4): `aria-label` on the Reveal/Hide button and `aria-live`
on the hidden/revealed status text.
- Security tripwire comment near the `saveSession` listener (finding #6):
captured auth bodies are unmasked at source (masking is UI-only), so any new
post-token-exchange persistence path must redact first.
Tests: form-encoded masking (token + refresh requests), JSON `code` NOT masked,
non-object JSON pass-through, default-fetch wrapper. Story play + NetworkEntry
tests updated for the new aria-labels and the now-masked form request body.
Finding #5 (pretty-print asymmetry) is already handled by ContentViewer's JSON
formatting; #4-aria and #6 are the doc/a11y bits.
Also merges the squash-merged #1385 base from v2/main.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* refactor(network): address #1387 third-pass review — content-type masking, wholesale mask, dedup (#1386)
Third-pass code-review feedback on the auth-body masking PR:
- maskSecretsInBody now takes the body's content-type (finding #1): `*json*`
→ JSON masking, form-urlencoded → form masking, any other known type
(HTML/plaintext/XML) → no masking; absent/unknown → sniff as before. Removes
the implicit "non-JSON ⇒ form" guess for error pages etc. NetworkEntry passes
the request/response `content-type` header through to BodyPreview.
- Mask any non-empty value under a sensitive key wholesale (finding #2):
a non-standard object/array value under e.g. `access_token` is replaced
rather than recursed into, so it can't leak. Empty-string still not flagged.
- Extracted `isSensitiveKey(set, key)` to dedupe the JSON/form key checks
(finding #3).
- Reworded `MaskResult.masked` doc to cover the form (non-pretty-printed) case
(finding #6).
Tests: non-string-value-under-sensitive-key wholesale mask; repeated form
param; empty form value not flagged; content-type honored (JSON type skips
form masking, text/plain skips masking, explicit form type masks).
Finding #5 (story re-hide) skipped per reviewer — unit test already covers
re-mask. Finding #4 (form edge cases) added.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* test(network): address #1387 fourth-pass nits — content-type contract, key, form test (#1386)
Fourth-pass review (LGTM) — minor, non-blocking items:
- Documented the content-type matching contract on `maskSecretsInBody`:
substring match (`*json*` / `*x-www-form-urlencoded*`), and we trust the
wire's own label (a mislabeled body renders raw — acceptable for the
screen-share threat model) (finding #1).
- Key `<BodyPreview>` by content-type + body (not body alone) so a header-only
change would also reset the reveal state (finding #3; not reachable today).
- Storybook `AuthSuccess`: assert revealed count `>=` reveal-button count to
mirror the `hidden >= 2` check, so adding a non-masked body later can't drift
the assertion silently (finding #4).
- Added a `NetworkEntry` unit test for form-encoded request-body masking
(code/code_verifier) — previously exercised only via the Storybook play.
Finding #2 (silent decodeURIComponent fallback on malformed keys) acknowledged
as a known quirk outside the threat model; no change.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* perf(network): address #1387 fifth-pass — memoize masking, mask DCR mgmt token (#1386)
Fifth-pass review (LGTM):
- Memoized `maskSecretsInBody` in `BodyPreview` (recommended item #1): a
Reveal/Hide click no longer re-parses + re-walks the body; cost is once per
mount, and the `key` remount re-runs it on body/content-type change. Guarded
so a too-large body is never parsed (the hook runs unconditionally, with the
size check inside the memo).
- Added `registration_access_token` (RFC 7592 DCR management credential, same
bearer class) to the masked key set, and a confidential-client `/register`
response fixture test (#5) asserting client_secret + registration_access_token
masked, client_id/metadata visible.
- Tightened the `isMaskableValue` doc to state the exact contract: non-null,
non-empty-string values are masked wholesale (#3).
Items #2 (transport parse cost — capped by the memo) and #4 (test-only
MASK_PLACEHOLDER export) left as-is per the review.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* docs(network): clarify body-update emits list event only (#1387 sixth pass) (#1386)
Sixth-pass review (LGTM, all nits). Added a comment on
`onFetchRequestBodyUpdate` noting it re-emits only `fetchRequestsChange`, not
the per-entry `fetchRequest` event — list-reading consumers pick up the body
on the next render; a future per-entry subscriber would need its own event.
Other items left as-is per the review: full-string `BodyPreview` key (#1,
micro-perf only if profiled), NaN under a sensitive key (#3, consistent with
the mask-non-null-non-empty contract), `&`-only form separator (#4, OAuth
never emits `;`).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(auth): inject MCP_INSPECTOR_API_TOKEN into served index.html (#1378)
Reloading the web client at the bare URL (no `?MCP_INSPECTOR_API_TOKEN=…`
query string) with empty sessionStorage made every `/api/*` request 401 —
the browser had no way to recover the backend's auth token.
Embed the token into `index.html` on every page load so the browser no
longer depends on the query string surviving navigation:
- New shared helper `clients/web/server/inject-auth-token.ts` embeds
`<script>window.__INSPECTOR_API_TOKEN__ = "…"</script>` (escaped against
`</script>` injection; no-op when auth is dangerously omitted).
- Dev: the Vite plugin injects via `transformIndexHtml`.
- Prod: the Hono server injects on the `/` route.
- `App.tsx` `getAuthToken()` now reads the injected global first, then the
query string, then sessionStorage (both fallbacks preserved).
- Shared global name lives in `INSPECTOR_API_TOKEN_GLOBAL`
(`core/mcp/remote/constants.ts`).
Tests: helper unit coverage + an integration test exercising the real
prod server's `/` → `/api/*` flow (injected token authenticates; missing
token 401s). AGENTS.md documents the token-recovery order.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(auth): wire OAuth authorization-code flow into App.tsx (#1379)
OAuth-protected MCP servers could not be connected to from the v2 web
client: the core OAuth pipeline exists, but App.tsx never invoked it, so a
connect attempt 401'd and surfaced "Remote send failed (401): … Missing
Authorization header" as a toast.
Wire the two missing entry points (all core primitives already in place):
- Auto-trigger on 401: in onToggleConnection's catch, detect an upstream
401 (isUnauthorizedError) and call client.authenticate(), which runs
discovery + DCR (backend-proxied) and redirects the page to the auth
server via BrowserNavigation. The initiating server id is persisted to
sessionStorage first, since the OAuth `state` carries only mode+authId
and the full-page redirect wipes React state.
- /oauth/callback handler: a mount effect that, once `servers` hydrate,
parses the callback params, recovers the pending server, rebuilds its
InspectorClient, runs completeOAuthFlow(code) (PKCE verifier + DCR client
info survive in BrowserOAuthStorage), replaceState("/") so a reload can't
replay the single-use code, then connect(). An `error=` callback toasts
instead of retrying.
connect() already attaches the OAuth provider to the transport
(inspectorClient.ts), so once tokens land in BrowserOAuthStorage the
outbound request carries the bearer token.
Extracted the pure pieces (constants + isUnauthorizedError) to
src/utils/oauthFlow.ts with unit tests. Verified end-to-end in a real
browser against the MCP SDK demo OAuth server: Connect -> redirect ->
auto-approve -> callback -> Connected, with the access token shown in the
Connection Info modal (#1377).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(auth): persist OAuth pre-redirect Network log across the redirect (#1384)
After #1379, the Network tab showed only the post-redirect auth HTTP
(discovery re-run + POST /token); the pre-redirect discovery and the DCR
POST /register that run during authenticate() were lost when the page
navigated to the auth server.
Root causes:
1. Ordering — BrowserNavigation set `window.location.href` before the
client's `saveSession` event fired (OAuthManager calls onBeforeOAuthRedirect
*after* auth() already navigated), so the save raced the unload and was
dropped. Fix: BrowserNavigation now runs a synchronous `beforeNavigate`
hook immediately before assigning location.href; App wires it through
createWebEnvironment to flush the active fetch log to RemoteInspectorClient
Storage (keyed by the authId parsed from the auth URL) via a keepalive POST
that outlives the unload.
2. Illegal invocation — RemoteInspectorClientStorage defaulted to
`this.fetchFn = globalThis.fetch` and called `this.fetchFn(...)`, which
re-binds `this` and makes native fetch throw "Illegal invocation"
(swallowed by the catch). This silently broke *all* session save/load.
Fix: default to a wrapper that preserves the global receiver.
3. Restore race — hydrateFetchRequests replaced the list, so a load that
resolved after the resuming connect appended live entries would clobber
them. Fix: merge restored (older) entries ahead of live ones, dedupe by id.
saveSession also now uses keepalive: true.
Verified end-to-end against the MCP SDK demo OAuth server: the connected
page's Network tab shows the full handshake — pre-redirect discovery + DCR
/register plus post-redirect discovery + /token as `auth`, alongside
`transport`. Added unit tests for the beforeNavigate ordering and the
hydrate merge/dedupe.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(auth): inject token into prod SPA fallback + prime sessionStorage + no-store (#1378)
Addresses code-review feedback on the token-injection PR:
- Prod `server.ts` SPA deep-link fallback (e.g. `/oauth/callback`) previously
served the raw index.html off disk via serveStatic, bypassing injection — so
a bookmark/reload at a non-`/` route with empty sessionStorage would 401.
Route the SPA fallback through the same `serveIndexHtml` (inject) helper;
real static assets (paths with a dot) still serve verbatim. Dev already
injected on every HTML serve via Vite `transformIndexHtml`.
- `getAuthToken()` now persists the injected `window.__INSPECTOR_API_TOKEN__`
to sessionStorage (not just the URL-param branch), priming the backstop for
any later navigation that loses the global.
- Injected HTML responses now send `Cache-Control: no-store`, so a page
carrying a token isn't cached and served stale after a restart regenerates
the token.
Integration tests added: SPA fallback (`/oauth/callback`) carries the token,
`Cache-Control: no-store` on injected HTML, real assets served verbatim, and
unknown `/api` routes 404 rather than falling through to the HTML shell.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(auth): address #1383 review — split OAuth/connect toasts, tighten 401 match (#1379)
Code-review feedback on the OAuth-wiring PR:
- Callback effect: split completeOAuthFlow vs connect() into separate
try/catch blocks. A token-exchange failure now reads "OAuth token exchange
failed … Please try connecting again." (the single-use code is spent and the
URL was cleared, so a reload can't retry); a post-OAuth connect failure reads
"Failed to connect" since OAuth actually succeeded and re-clicking Connect
reuses the persisted tokens.
- isUnauthorizedError: anchor the message fallback on the transport's
`failed …(401)` wording instead of a bare `(401)`, so an unrelated `(401)`
spliced into an error message can't trip the OAuth flow. Added a test.
- Documented that clearing the pending id + URL before the server lookup is
intentional (deleted/renamed server mid-flow → require a fresh Connect).
Also merges the squash-merged #1382 base from v2/main.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* docs(auth): address #1385 review — clarify double-save + keepalive cap + test fetch default (#1384)
Code-review feedback on the OAuth Network-log persistence PR:
- Documented the double-save: `FetchRequestLogState`'s `saveSession` listener
is the backstop; `BrowserNavigation`'s `beforeNavigate` hook is the primary
flush for the redirect case. Notes the listener may lose the navigation race
and is harmless when it duplicates (last-writer-wins, identical payload).
- Reworded the keepalive comment in `RemoteInspectorClientStorage.saveSession`:
the 64KB cap is general (the method is also reachable from the listener with
the full session log), so a long session could exceed it and drop silently —
acceptable since the persisted log is best-effort, not load-bearing.
- Added a regression test that constructs `RemoteInspectorClientStorage`
without a `fetchFn`, stubs `globalThis.fetch`, and asserts the default
wrapper calls it (locks in the "Illegal invocation" fix, which callers
otherwise swallow).
Optional items (logger.warn on swallowed save errors; setupClientForServer dep
churn) acknowledged on the PR, not changed.
Also merges the squash-merged #1383 base from v2/main.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(auth): inject MCP_INSPECTOR_API_TOKEN into served index.html (#1378)
Reloading the web client at the bare URL (no `?MCP_INSPECTOR_API_TOKEN=…`
query string) with empty sessionStorage made every `/api/*` request 401 —
the browser had no way to recover the backend's auth token.
Embed the token into `index.html` on every page load so the browser no
longer depends on the query string surviving navigation:
- New shared helper `clients/web/server/inject-auth-token.ts` embeds
`<script>window.__INSPECTOR_API_TOKEN__ = "…"</script>` (escaped against
`</script>` injection; no-op when auth is dangerously omitted).
- Dev: the Vite plugin injects via `transformIndexHtml`.
- Prod: the Hono server injects on the `/` route.
- `App.tsx` `getAuthToken()` now reads the injected global first, then the
query string, then sessionStorage (both fallbacks preserved).
- Shared global name lives in `INSPECTOR_API_TOKEN_GLOBAL`
(`core/mcp/remote/constants.ts`).
Tests: helper unit coverage + an integration test exercising the real
prod server's `/` → `/api/*` flow (injected token authenticates; missing
token 401s). AGENTS.md documents the token-recovery order.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(auth): wire OAuth authorization-code flow into App.tsx (#1379)
OAuth-protected MCP servers could not be connected to from the v2 web
client: the core OAuth pipeline exists, but App.tsx never invoked it, so a
connect attempt 401'd and surfaced "Remote send failed (401): … Missing
Authorization header" as a toast.
Wire the two missing entry points (all core primitives already in place):
- Auto-trigger on 401: in onToggleConnection's catch, detect an upstream
401 (isUnauthorizedError) and call client.authenticate(), which runs
discovery + DCR (backend-proxied) and redirects the page to the auth
server via BrowserNavigation. The initiating server id is persisted to
sessionStorage first, since the OAuth `state` carries only mode+authId
and the full-page redirect wipes React state.
- /oauth/callback handler: a mount effect that, once `servers` hydrate,
parses the callback params, recovers the pending server, rebuilds its
InspectorClient, runs completeOAuthFlow(code) (PKCE verifier + DCR client
info survive in BrowserOAuthStorage), replaceState("/") so a reload can't
replay the single-use code, then connect(). An `error=` callback toasts
instead of retrying.
connect() already attaches the OAuth provider to the transport
(inspectorClient.ts), so once tokens land in BrowserOAuthStorage the
outbound request carries the bearer token.
Extracted the pure pieces (constants + isUnauthorizedError) to
src/utils/oauthFlow.ts with unit tests. Verified end-to-end in a real
browser against the MCP SDK demo OAuth server: Connect -> redirect ->
auto-approve -> callback -> Connected, with the access token shown in the
Connection Info modal (#1377).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(auth): inject token into prod SPA fallback + prime sessionStorage + no-store (#1378)
Addresses code-review feedback on the token-injection PR:
- Prod `server.ts` SPA deep-link fallback (e.g. `/oauth/callback`) previously
served the raw index.html off disk via serveStatic, bypassing injection — so
a bookmark/reload at a non-`/` route with empty sessionStorage would 401.
Route the SPA fallback through the same `serveIndexHtml` (inject) helper;
real static assets (paths with a dot) still serve verbatim. Dev already
injected on every HTML serve via Vite `transformIndexHtml`.
- `getAuthToken()` now persists the injected `window.__INSPECTOR_API_TOKEN__`
to sessionStorage (not just the URL-param branch), priming the backstop for
any later navigation that loses the global.
- Injected HTML responses now send `Cache-Control: no-store`, so a page
carrying a token isn't cached and served stale after a restart regenerates
the token.
Integration tests added: SPA fallback (`/oauth/callback`) carries the token,
`Cache-Control: no-store` on injected HTML, real assets served verbatim, and
unknown `/api` routes 404 rather than falling through to the HTML shell.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(auth): address #1383 review — split OAuth/connect toasts, tighten 401 match (#1379)
Code-review feedback on the OAuth-wiring PR:
- Callback effect: split completeOAuthFlow vs connect() into separate
try/catch blocks. A token-exchange failure now reads "OAuth token exchange
failed … Please try connecting again." (the single-use code is spent and the
URL was cleared, so a reload can't retry); a post-OAuth connect failure reads
"Failed to connect" since OAuth actually succeeded and re-clicking Connect
reuses the persisted tokens.
- isUnauthorizedError: anchor the message fallback on the transport's
`failed …(401)` wording instead of a bare `(401)`, so an unrelated `(401)`
spliced into an error message can't trip the OAuth flow. Added a test.
- Documented that clearing the pending id + URL before the server lookup is
intentional (deleted/renamed server mid-flow → require a fresh Connect).
Also merges the squash-merged #1382 base from v2/main.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(auth): inject MCP_INSPECTOR_API_TOKEN into served index.html (#1378)
Reloading the web client at the bare URL (no `?MCP_INSPECTOR_API_TOKEN=…`
query string) with empty sessionStorage made every `/api/*` request 401 —
the browser had no way to recover the backend's auth token.
Embed the token into `index.html` on every page load so the browser no
longer depends on the query string surviving navigation:
- New shared helper `clients/web/server/inject-auth-token.ts` embeds
`<script>window.__INSPECTOR_API_TOKEN__ = "…"</script>` (escaped against
`</script>` injection; no-op when auth is dangerously omitted).
- Dev: the Vite plugin injects via `transformIndexHtml`.
- Prod: the Hono server injects on the `/` route.
- `App.tsx` `getAuthToken()` now reads the injected global first, then the
query string, then sessionStorage (both fallbacks preserved).
- Shared global name lives in `INSPECTOR_API_TOKEN_GLOBAL`
(`core/mcp/remote/constants.ts`).
Tests: helper unit coverage + an integration test exercising the real
prod server's `/` → `/api/*` flow (injected token authenticates; missing
token 401s). AGENTS.md documents the token-recovery order.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(auth): inject token into prod SPA fallback + prime sessionStorage + no-store (#1378)
Addresses code-review feedback on the token-injection PR:
- Prod `server.ts` SPA deep-link fallback (e.g. `/oauth/callback`) previously
served the raw index.html off disk via serveStatic, bypassing injection — so
a bookmark/reload at a non-`/` route with empty sessionStorage would 401.
Route the SPA fallback through the same `serveIndexHtml` (inject) helper;
real static assets (paths with a dot) still serve verbatim. Dev already
injected on every HTML serve via Vite `transformIndexHtml`.
- `getAuthToken()` now persists the injected `window.__INSPECTOR_API_TOKEN__`
to sessionStorage (not just the URL-param branch), priming the backstop for
any later navigation that loses the global.
- Injected HTML responses now send `Cache-Control: no-store`, so a page
carrying a token isn't cached and served stale after a restart regenerates
the token.
Integration tests added: SPA fallback (`/oauth/callback`) carries the token,
`Cache-Control: no-store` on injected HTML, real assets served verbatim, and
unknown `/api` routes 404 rather than falling through to the HTML shell.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(server-info): wire Server Info modal to the connection handshake
Closes#1364
- Add ServerInfoModal wrapping the existing ServerInfoContent, mirroring the
ServerSettingsModal layout.
- Gate the Server Info button on ServerCard to render only when the
connection is "connected".
- Expose `getClientCapabilities()` on InspectorClient (and protocol +
FakeInspectorClient), and surface it through useInspectorClient so the
modal can render the client side of the capability handshake without
poking the SDK Client's private state.
- App.tsx: open the modal from the gated Server Info button, derive the
active server's transport and OAuth details (synchronously, from the
guided-OAuth state snapshot + persisted settings), and reset the open
flag on InspectorClient `disconnect` so the modal does not auto-reopen
on a future reconnect.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(server-info): dedupe modal header and render instructions in ContentViewer
- Drop the in-content "Server Information" Title from ServerInfoContent and
rename the modal title to "Server Information" so the modal renders one
heading instead of two stacked ones.
- Swap the instructions Blockquote for ContentViewer (copyable, with the
existing wrapping Code variant) so long instructions wrap and the user
can copy the full payload.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* refactor(server-info): rename ServerInfo -> ConnectionInfo
The popup covers server identity AND client capabilities AND OAuth state —
it's a snapshot of the live connection, not just the server. Rename to match
what it actually shows. Also add a "Server Implementation" heading above
the top section (name/version/protocol/transport) to balance the existing
capability headings.
- ServerInfoContent -> ConnectionInfoContent (directory, file, exports)
- ServerInfoModal -> ConnectionInfoModal (directory, file, exports)
- ServerCard button label "Server Info" -> "Connection Info";
prop onServerInfo -> onConnectionInfo
- Modal title -> "Connection Info"
- App.tsx / ServerListScreen / InspectorView wiring updated to match
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(connection-info): cap instructions block height so it scrolls in place
A long server prompt (e.g. an Everything-style instruction document) was
pushing the OAuth section and the modal chrome off-screen — users had to
scroll the whole modal to reach the close button. Wrap the ContentViewer
in ScrollArea.Autosize with mah=280 so the instructions section scrolls
internally and the rest of the modal stays put.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat(connection-info): add subtle background to section headings
Add a `section` variant for Title that paints the heading in the existing
`--inspector-surface-subtle` token (same surface token already used for
modal call-out panels). Apply it to each section heading in
ConnectionInfoContent (Server Implementation, Server Capabilities, Client
Capabilities, Server Instructions, OAuth Details) via a SectionHeading
subcomponent constant, so the sections read as distinct cards against the
modal background instead of bleeding into one stream of text.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(stories): wrap long import line in ConnectionInfoModal.stories.tsx
Prettier flagged the import in CI. The line exceeded printWidth=80 and
needed to wrap onto multiple lines.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* docs(agents): require `npm run format` before committing
Add a pre-commit instruction to run the auto-fixing `npm run format` ahead
of `npm run validate`. `validate` runs `format:check` (non-fixing) which
will fail CI on any unformatted file — running the auto-fixer first
eliminates that class of failure entirely.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* chore(favicon): use theme/assets/MCP.svg as the favicon source of truth
Point index.html's icon link at /src/theme/assets/MCP.svg so the same MCP
logo asset already imported by ViewHeader is also the favicon. Delete the
now-unused clients/web/public/favicon.svg. Vite resolves the source path
during build and emits a hashed copy under dist/assets/.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(server-card): surface failed-handshake errors with a 5s auto-dismiss
Two related fixes so the user actually sees what went wrong when a
connect attempt fails (e.g. OAuth-backed server, bad URL, missing dev
auth token):
- App.tsx: add `connectInFlightRef` and skip clearing `activeServerId`
in the InspectorClient `disconnect` listener while a connect attempt
is in flight. The transport closes during a failing handshake and
dispatches `disconnect` before the catch block sets `errorMessage`;
without this guard InspectorView would map the card to plain
"disconnected" and the InlineError would never render.
- InlineError: opt-in `autoDismissMs` prop that runs a Mantine `slide-up`
Transition exit animation after the message has been visible for the
given duration. Visibility is derived from a "dismissed message"
state (rather than a mounted boolean + setMounted-in-effect), so the
timer re-arms whenever `error.message` changes without tripping the
react-hooks/set-state-in-effect rule. ServerCard passes
autoDismissMs={5000}.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* refactor(server-card): own InlineError visibility + animate parent-driven removal
Move the 5s auto-dismiss timer and visibility state out of InlineError
and into ServerCard so the slide-up exit animation runs in both cases:
- Timer-driven: after 5s the alert animates out (existing behavior).
- Parent-driven: when `connection.error` is cleared mid-flight (e.g. a
successful reconnect), the alert ALSO slides up instead of vanishing.
To keep the message painted during the parent-driven exit, ServerCard
tracks a `lastError` state that mirrors `connection.error` only on its
truthy edges. The conditional setState during render is the supported
React pattern for "remember the last seen value of a prop"; gated on
message inequality so it can't loop. The `useRef` version of this
pattern failed `react-hooks/refs` (no ref writes/reads during render).
InlineError is now a pure presentational component: `mounted` prop
controls a Mantine `slide-up` Transition. The internal `autoDismissMs`
prop is gone — its single caller has moved to the controlled `mounted`
pattern.
Closes the "error vanishes the moment the parent removes it" half of #1364
(the timer half landed in 6d147e0f).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* refactor(server-card): surface connection errors as a toast, drop InlineError
Move handshake-failure messaging out of the ServerCard body and onto a
Mantine toast (`notifications.show`). The inline alert had a 5-second
visibility contract entangled with InspectorView's status pipeline, and
the active card was flipping through error → disconnected → error during
a failing handshake; surfacing the error as a toast sidesteps that whole
timing problem and keeps the card focused on the live status indicator.
- App.tsx: toast on the connect-catch path; drop `errorMessage` state +
`connectInFlightRef` (the latter was only there to keep the inline
alert visible past the transport's close event).
- InspectorView: drop the `errorMessage` prop and the
`connectionStatus === "error" && errorMessage` mapping branch.
- ServerCard: drop the InlineError rendering + the displayedError /
errorVisible visibility state machine.
- Delete the InlineError component — no other consumers.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* chore(notifications): anchor toasts to the bottom right
Move the Notifications portal from top-right to bottom-right so a
connection-failure toast doesn't overlap the ViewHeader / tabs row.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* review: address PR feedback on the Connection Info modal
1. Hoist `EMPTY_CLIENT_CAPABILITIES = Object.freeze({})` to module scope
in `useInspectorClient` so the fallback returns a stable reference
(avoids invalidating downstream `useMemo`/`useEffect` deps when no
client is attached).
4. Drop the `?? "stdio"` silent fallback on `connectionInfoTransport`
in the JSX — gate the modal render on `activeServer` too. The
"stdio" default still lives at the derivation site (config.type is
schema-optional and `ServerType` is required by the modal prop),
but the gate makes the invariant explicit.
6. Replace the raw `document.querySelector("button.mantine-CloseButton-root")`
in `ConnectionInfoModal.test.tsx` with `screen.getByRole("button",
{ name: /close/i })`. Add `aria-label="Close"` to the
CloseButton so the assertion (and screen readers) have a name to
match on.
7. Switch the OAuth scope splitter from `/\s+/` to the literal `" "`
per OAuth 2.1 §3.3.
Also clarify in comments:
2. Note that `connectionInfoOAuth` uses snapshot-at-open semantics on
purpose — token refreshes mid-session won't re-render the modal.
5. Note that we prefer the AS-advertised `authorization_endpoint`
over `authorizationUrl` because the latter is per-flight noise
(state, code_challenge) rather than a stable identifier.
Finding 3 (modal exit animation short-circuited by the `{initializeResult
&& …}` guard) left as intentional — we want the snap-shut on disconnect
so a stale info dialog can't keep rendering with no live client behind it.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* docs(connection-info): tighten the OAuth memo comment per re-review
The reviewer noted that "snapshot-at-open" was imprecise — the memo
re-runs on any of its deps (connectionStatus, inspectorClient,
activeServer), which includes settings edits that re-reference the
active server. Reframe as "snapshot-at-last-derivation" and call out
the natural triggers (server switch, reconnect, settings edit) so a
future reader doesn't conclude the memo is open-only.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat(sort): add SortToggle for Logs, History, and Network (#1371)
Introduces a reusable SortToggle element and wires it into the Logs,
History, and Network screens so users can flip chronological order.
Selection is persisted per-screen to localStorage under the
inspector.sortDirection.* namespace, with invalid stored values falling
back to the newest-first default.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(logs): move auto-scroll checkbox left of the sort selector
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* refactor(logs): remove auto-scroll checkbox; ListToggle left of SortToggle
- Drops the Logs auto-scroll checkbox and its plumbing through
LoggingScreen / InspectorView.
- Moves the ListToggle to the left of the SortToggle on the History
and Network panel toolbars.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* refactor(sort): address review on #1372
- Hydrate sort direction synchronously (getInitialValueInEffect: false)
to remove the one-frame newest-first flicker on first paint.
- Drop the redundant `[...matched].sort()` copy across the three panels;
`.filter()` already returns a fresh array.
- Document why serializeSortDirection bypasses Mantine's default JSON
stringify (raw enum value in storage).
- MemoryStorage.getItem: collapse `has` + `!` to `?? null`.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(history): default the request list to collapsed for parity with Network
The History list opened expanded while Network opened collapsed; align
History with Network so both panels start in the same compact view.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat(history,network): persist list compact state to localStorage
Lifts the per-panel `compact` flag out of HistoryListPanel /
NetworkStreamPanel and into InspectorView's localStorage hooks, mirroring
the sort-direction shape:
inspector.listCompact.history
inspector.listCompact.network
Defaults to collapsed (compact=true) on both, with a corrupted-value
deserializer that clamps unknown strings back to the default.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat(servers,resources): persist ListToggle state to localStorage
Lifts the per-screen compact preference out of ServerListScreen and
ResourceControls into InspectorView's localStorage hooks, matching the
History/Network shape:
inspector.listCompact.servers
inspector.listCompact.resources
Servers and Resources both default to expanded (compact=false). For
ResourceControls the displayed compact is still derived from openSections
(so per-section accordion clicks still work locally), and only explicit
ListToggle clicks update the persisted preference — the handler computes
the next compact value from the click outcome rather than blindly
flipping, so a half-open accordion still records the right post-state.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* refactor(sort-toggle): switch to icon button matching ListToggle style
Replaces the Mantine Select with a subtle Button (or ActionIcon for the
subtle variant), using TbSortDescending2 for newest-first and
TbSortAscending2 for oldest-first. Clicking the button flips between the
two values.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* refactor(sort-toggle): restore Select with sort icon in rightSection
Reverts SortToggle to a Mantine Select with simplified labels ("Newest
First" / "Oldest First") and the corresponding TbSortDescending2 /
TbSortAscending2 icon in the rightSection, replacing the default chevron.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(sort-toggle): swap icons so they match label direction
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(sort-toggle): use down arrow for newest, up arrow for oldest
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* refactor(inspector-view): extract useSortDirection / useListCompact
Collapses seven near-identical useLocalStorage call sites in
InspectorView into two scoped helpers. Each helper carries the shared
namespace, serialize/deserialize adapters, and getInitialValueInEffect:
false; the call sites now read as "this screen has sort persisted to
this scope" rather than re-stating the storage shape every time.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(sort-toggle): apply prettier formatting
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat(export): wire History + Logs export JSON handlers (#1354)
Replaces the `todoNoop` placeholders in App.tsx with real handlers that
serialize the current entries and trigger a browser download via the
existing `downloadJsonFile` helper.
- Adds `buildExportFilename(kind, serverId, now?)` next to the download
helper. Produces sortable, filesystem-safe names like
`inspector-history-alpha-2026-03-17T10-00-42.123Z.json` (URL-encodes
server ids that contain slashes / spaces / colons, swaps `:` for `-`
in the ISO timestamp so the file is safe on Windows).
- Wires `onExportHistory` (full MessageEntry list) and `onExportLogs`
(the LogEntryData list the screen renders).
- Routes the existing Network export through the same filename helper
for consistency.
Unit tests cover the filename builder, including the no-server-id
fallback, special-character encoding, and the default-`now` shape.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(export): tighten buildExportFilename signature + docs
- Introduce an `ExportKind = "history" | "logs" | "network"` union and
use it as the type of the `kind` parameter. Typos like "hisotry" now
fail at compile time and the supported export kinds are documented
inline.
- Clarify the JSDoc: the segment is omitted when `serverId` is falsy
(undefined OR empty string), matching the existing `serverId ? ... :
undefined` check. New test pins the empty-string branch.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* chore(logging): remove Copy All button from LogStreamPanel
Export now produces the same JSON the Copy All button used to dump to
the clipboard, so the button is redundant. History and Network panels
have no equivalent — dropping Copy All also brings Logs into parity.
Removes the prop from LogStreamPanel + LoggingScreen + InspectorView +
App.tsx, and updates the matching tests/stories. Tests now assert the
button is *not* in the document so a future re-introduction would have
to update the assertion deliberately.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat(network): add Network screen surfacing backend fetch log (#1355)
Routes `FetchRequestLogState` to a new Network tab so users can inspect
HTTP-layer request/response headers, status, duration, and category
(auth vs transport) of every fetch the backend makes for an active
session. Adds `NetworkControls` + `NetworkStreamPanel` + `NetworkEntry`
groups, a `monoBreak` Text theme variant for long header values, and
wires onClearNetwork / onExportNetwork (JSON download) from `App.tsx`.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(network): include request / response bodies in search filter
Headers were already covered but bodies were not, so searches for
strings that live inside the JSON-RPC payload (e.g. "jsonrpc",
"initialize", a tool name) silently returned no matches.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(network): always render Response Body section, note streaming
Streaming responses (text/event-stream, application/x-ndjson, POST to
/mcp) are intentionally not body-captured by the fetch tracker, so the
section was silently omitted. Now whenever a response was received we
render the section, with either the body, a "(empty)" note, or a
"Streaming response — body not captured" note keyed off content-type.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(network): capture response bodies via async body-update events
Two related fixes:
1. The fetch tracker was marking every streamable-HTTP POST /mcp call as
"not captured" because the response content-type is text/event-stream.
In practice those streams are bounded — the server sends the
JSON-RPC reply (sometimes preceded by progress events) and closes —
so we can clone the body and read it. Only GET + SSE / ndjson is the
unbounded long-lived channel that has to stay uncaptured.
2. Reading the body inline blocked the transport from processing
progress notifications until the entire stream finished, breaking the
`resetTimeoutOnProgress` integration test. The body is now read in
the background and dispatched via a new `fetchRequestBodyUpdate`
event that the FetchRequestLogState patches into the matching entry
by id. Plumbing wires through node/transport, the remote SSE channel
(RemoteSession, RemoteClientTransport), and the remote event types.
Also hides the Network tab when the active server is stdio (no HTTP
traffic to surface) and bumps the inline body preview limit to 100 KB
so typical tools/list responses render rather than tripping the "too
large" notice.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat(network): show orange SSE badge for long-lived stream entries
Adds a small "SSE" badge to the right side of the NetworkEntry header
(left of the HTTP status badge) when the response is a long-lived
streaming channel (GET + text/event-stream or application/x-ndjson) —
the only case where the body genuinely isn't captured. Makes the
"not captured" placeholder less surprising at a glance.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(network): address PR review findings
- Stop capturing OAuth response bodies (drop updateResponseBody from the
effectiveAuthFetch tracker). Restores pre-PR behavior for auth-category
entries so live access_token / refresh_token values can't leak into the
Network tab body preview during a screen-share. Headers + status are
still tracked.
- Extract `isLongLivedStreamResponse(method, contentType)` to
core/mcp/fetchTracking so the rule lives in one place; consumed by the
tracker and the NetworkEntry UI placeholder.
- Document the intent of NetworkEntry's `useEffect` (the list-level
toggle is authoritative; per-entry overrides are intentionally
discarded on each list toggle).
- Rename MAX_INLINE_BODY_BYTES → MAX_INLINE_BODY_CHARS (`.length` is
UTF-16 code units, not bytes) and update the user-facing notice.
- Switch NetworkStreamPanel's search from joined-string `.includes` to
per-field `.some(.includes)` so a term can't span field boundaries.
Regression test added.
- Add aria-label to ListToggle ("Expand all" / "Collapse all") — both an
a11y improvement and a stable selector that lets the panel test stop
fishing the button out by text-content negation.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* test(network): use ListToggle aria-label in panel + modal tests
Cleanup spotted in PR review follow-up:
- HistoryListPanel + ServerSettingsModal previously fished the ListToggle
out by empty text content because the toggle had no accessible name.
Now that ListToggle exposes "Expand all" / "Collapse all", both tests
can use getByRole + name and the stale comments go away.
- ListToggle's own tests now assert the aria-label contract so a future
refactor can't silently drop it.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(network): expose category toggle state via aria-pressed
The category UnstyledButtons in NetworkControls previously communicated
their toggled state only via background color. Adds aria-pressed={active}
so assistive tech can discover the on/off state without visuals, with a
test that asserts the attribute tracks the visibleCategories prop.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The four ServerConfigModal play functions used `findByText("Add server")`
etc. to assert the modal mounted. Mantine's Modal portals to document.body
outside the storybook canvas, and its first paint can race against the
1000ms default that `findBy*` queries retry against — especially in the
Storybook dev UI (slower than `test:storybook`'s headless browser,
particularly with devtools open or coverage instrumentation on). When the
play fired before the modal mounted, findByText timed out, Storybook
flagged the story as FAIL, and left the "preparing" loader visible.
Switch to `findByRole('dialog', { name })` with a 5s timeout, and scope
subsequent queries (`getByLabelText`, etc.) to the dialog's subtree via
`within(dialog)`. The role-based query asserts the dialog actually mounted
(more semantically meaningful than a free text lookup), the longer ceiling
covers slow first-paint, and scoping to the dialog avoids false-positive
matches from sibling loader overlays.
Verified locally: 1758 tests pass, 8/8 consecutive reloads of the Add Empty
story in the Storybook dev UI now show "Pass".
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(servers): make settings form re-render on every keystroke (#1361)
The settings modal is fully controlled by App.tsx, but pending edits were
stashed in a useRef — refs don't trigger re-renders, so the input's `value`
prop stayed stale until the 300ms debounce + PUT + refetch round-tripped.
Clicking "Add Header" / "Add Metadata" showed no row, and typing in OAuth
fields had no character echo, until that round-trip completed (~500ms).
The bug landed in #1353 with the debounced-flush pattern. It surfaced after
#1356 because per-keystroke PUTs now also write the keychain, making the
round-trip latency more noticeable.
Lift the draft into `useState`. Every change re-renders the modal
immediately. Keep the 300ms debounce on the PUT side. Initialize the draft
from the server-list entry only when the modal opens to a new target (not
on every `servers` change) so background refreshes don't clobber the user's
in-progress edits. Flush on close to preserve final keystrokes.
Verified live against the dev server: header/metadata rows appear on click,
OAuth Client ID / Client Secret typing echoes without lag, and the values
persist correctly (clientId in mcp.json, clientSecret in the macOS
Keychain).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* chore(servers): extract useSettingsDraft hook + tests; review polish (#1361)
Picks up the three review observations on #1362:
1. Extract the draft state + debounce + flush logic from App.tsx into
`core/react/useSettingsDraft`. Same behavior; thinner App.tsx;
testable in isolation with `renderHook`. Tests pin the regression
#1361 was about (synchronous onChange → draft update) plus the
debounce window, burst-collapse, flush-on-close, error propagation,
and the "background refresh must not clobber in-progress edits"
contract via a `resolveInitial` closure swap.
2. Hoist the empty `InspectorServerSettings` fallback to a module-scope
`EMPTY_SETTINGS` const. Both the hook's init path and the modal's
closed-state fallback share the same object identity; the inline
`useMemo` that previously re-derived it is gone.
3. Add a comment on the `resolveInitial` callback noting that we read
`servers` from this render's closure under the assumption that the
entry point is the "Settings" button on a rendered server card.
Documents the contract for future callers (e.g. a keyboard shortcut
that opens the modal pre-hydration).
Validate clean: 1756 tests pass. New useSettingsDraft.test.tsx adds 13
test cases.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* chore(servers): stabilize flush identity + add unmount-cleanup test (#1361)
Two non-blocking observations from the third review on #1362:
1. `flush` identity churned on every keystroke because the useCallback
closed over `draft` + `targetId`. Read both through refs instead so
the returned callback identity is stable. The behavior is unchanged
— but the consumer's `onSettingsModalClose` (which wraps `flush`)
no longer re-allocates per keystroke, which means the modal's
`onClose` prop identity also stays stable. A new test pins this.
2. Added a comment in `onChange` calling out that callers switching
`targetId` mid-debounce must `flush()` first to preserve the
pending PUT. The only switch path today (`onSettingsModalClose` in
App.tsx) does that already — the comment is for future refactors.
3. New unmount-cleanup test pins the intentional "drop final
<debounceMs window on unmount" trade-off — flushing on unmount
could fire a PUT against an unmounting component, which is a
worse footgun than losing a few hundred ms of typing on route
change / HMR.
Validate clean: 144 files, 1758 tests pass.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat(servers): move oauthClientSecret and stdio env values to OS keychain (#1356)
mcp.json is designed to be tool-shareable (symlinked from Claude Desktop's
config, pasted into bug reports, synced via dotfiles), so storing OAuth
client secrets and stdio env values in plaintext there meant any of those
flows could leak them. Lift both into the OS keychain via @napi-rs/keyring
(active replacement for the archived keytar). The wire shape is unchanged:
the GET /api/servers handler rehydrates from the keychain so browser code
sees the same JSON; the on-disk file no longer contains the secret
material. Includes an idempotent migration that lifts plaintext from older
mcp.json files (or hand-edited ones) into the keychain on first read.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(servers): tolerate keychain unavailability on non-set ops (#1356)
CI on Linux failed because `@napi-rs/keyring` hard-errors without
libsecret, and the route handlers were touching the keychain even for
no-secret flows (defensive `deleteAllForServer` on POST/DELETE, sweep
on PUT). The user's intent for the Linux fallback was "hard fail when a
secret is actually involved" — not on every keychain touch.
Make `KeyringSecretStore.get / delete / deleteAllForServer` silently
degrade when the keychain is unavailable (return null / no-op); `set`
remains the one operation that hard-fails with
`KeychainUnavailableError`. The migration path catches the error and
preserves the disk plaintext rather than losing the secret.
Also inject `InMemorySecretStore` in `useServers.test.tsx` and
`servers-events.test.ts` so tests never touch a developer's real
keychain even when libsecret is present.
Adds `KeyringSecretStore` tests (vi.mock'd native bindings) and an
integration suite that simulates Linux-without-libsecret to verify the
tolerance contract end-to-end.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(servers): close GET migration race; reverse keychain/disk order (#1356)
Addresses code review:
1. **Concurrency**: GET /api/servers was reading + migrating outside the
write lock, then writing inside it — a concurrent POST/PUT/DELETE
between the read and the lock acquisition could be clobbered by the
migration write. Also affected the first-launch seed write. Fix:
take a fast unlocked peek; if the file is absent (seed branch) or
carries plaintext to migrate (migration branch), re-read inside the
write lock and decide there. Added `hasPlaintextSecrets()` as a
pure predicate so the common no-migration path stays lock-free.
2. **Write ordering**: POST and PUT wrote disk first, then keychain.
A keychain `set` failure mid-write left a half-configured entry on
disk and trapped retries at 409. Reversed to keychain-first so a
503 leaves both stores in their pre-write state. PUT in-place now
sets new fields, writes disk, then deletes obsolete fields (a
failed disk write leaves orphan keychain entries — recoverable —
rather than premature deletions of fields the user still expects
to see).
3. **Tolerance simplification**: dropped the no-entry regex in
`KeyringSecretStore.delete` — both "no entry" and "keychain
unavailable" collapse to the same desired outcome (the entry isn't
there anymore), so a uniform swallow is clearer.
4. **Parallelism**: `readKeychainEntriesFor` now `Promise.all`s the
per-field `get` calls. macOS Keychain round-trips are 10-50ms;
serializing 5+ env vars per server × N servers is a meaningful
wall-clock cost on GET.
5. **Spec doc**: reflects the tolerance contract (only `set`
hard-fails) and the keychain-before-disk write ordering rationale.
New regression test in servers-route.test.ts covers the retry-after-503
contract directly.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* chore(servers): parallel keychain writes + migration partial-state comment (#1356)
Two nice-to-have observations from the re-review:
- `writeKeychainEntriesFor` now `Promise.all`s its per-field sets,
matching the read side. No ordering requirement among distinct
(id, field) keys; on unavailability every set rejects and Promise.all
surfaces the first rejection (still 503 from the route handler).
- `migratePlaintextSecrets` catch block now spells out the
partial-migration semantics: we return the original config so the
disk plaintext stays intact, and the next successful GET retries
with the keychain-wins branch absorbing already-set entries.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* chore(servers): parallel keychain deletes for symmetry (#1356)
Picks up the symmetry observation from the third review: `deleteKeychainFields`
now uses `Promise.all` like its read and write siblings. Distinct (id, field)
deletes have no ordering requirement, and `secretStore.delete` is already a
silent no-op on unavailability so Promise.all has no failure-mode surprise.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat(servers): flatten mcp.json settings to top-level fields (#1358)
Eliminate the nested `settings` block introduced in #1352 so each
server entry's Inspector-extension fields live as direct keys alongside
`type` / `url`, matching the `.mcp.json` shape Claude Code / Cursor /
Cline write:
- `headers` becomes a `Record<string, string>` on disk (was nested
`settings.headers` as a pair-array)
- `oauth` becomes a nested `{ clientId?, clientSecret?, scopes? }`
object (was flat `settings.oauthClientId` / `oauthClientSecret` /
`oauthScopes`)
- `metadata`, `connectionTimeout`, `requestTimeout` move to the
entry's top level as-is (Inspector-only fields)
The in-memory + wire shape is unchanged from #1352: `InspectorServerSettings`
keeps pair-array `headers` and flat `oauth*` fields because the form
needs them in that shape to drive controlled-component editing. The
bidirectional conversion lives in `serverList.ts`
(`storedFieldsToInspectorSettings` / `inspectorSettingsToStoredFields`)
and is invoked by `mcpConfigToServerEntries` / `serverEntriesToMcpConfig`
and by the server route's `buildStoredEntry`. The `PUT /api/servers/:id`
wire envelope keeps the preserve/clear/apply semantics from #1353 — the
backend splats validated `settings.*` onto top-level disk keys when
assembling the next on-disk shape.
Hard-cutover migration per the issue's decision 4: files written by the
one pre-#1358 v2/main build have a nested `settings` block; we drop the
node on read with a one-line warn including the server id. The persisted
headers / metadata / timeouts / OAuth credentials are intentionally lost
on first read — users re-enter them via the form (or hand-edit the file
into the flat shape). v2 has not shipped a stable release with the
nested shape, so the blast radius is dogfooders only.
The smuggle-guard in `buildStoredEntry` widens from just `settings` to
all six Inspector-extension keys (`settings`, `headers`, `metadata`,
`connectionTimeout`, `requestTimeout`, `oauth`), so a caller can't
bypass `validateSettings` by nesting any of them inside `config`.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(servers): address PR review on flatten (#1359)
- Single-source the Inspector-extension key list with `satisfies`
(finding 1). `INSPECTOR_FIELD_KEY_MAP satisfies Record<keyof
StoredInspectorFields, true>` in serverList.ts means a new field on
the type forces a compile error here. The set is exported and
re-used by the server route's smuggle guard, so adding `proxy` or
similar in the future updates all three sites in lockstep. The PUT
preserve path now uses a new `stripInspectorFields` helper that
derives from the same source instead of hand-destructuring.
- Add read-path shape validation in `normalizeMcpServers` (finding 2):
`headers: "oops"` / `metadata: 42` / `oauth: []` / non-numeric
timeouts are dropped individually with a warn, so a hand-edited
malformed file can't put garbage rows into the form. Mirrors the
write-side `validateSettings` symmetry. New integration test covers
each malformed shape with a partial fixture so a single bad key
doesn't take out the whole entry.
- Align new warns to the `fileLogger || console.warn` fallback pattern
(finding 3) via a small `logWarn` helper, so the legacy-drop and
smuggle warns are visible in `npm run dev` where the logger is
often unconfigured.
- Spec doc tweak: the CLI/TUI out-of-scope note's "writes to
`settings.headers`" reference now reads "writes to the entry's
top-level `headers` field on disk (post-#1358 flat shape)" (finding
4). Avoids confusion for future readers.
- Nits: replace `current.mcpServers[originalId]!` with explicit
narrowing; add a comment on the OAuth truthiness drop in
`storedFieldsToInspectorSettings` mirroring the write-side comment.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(servers): address pass-2 review on flatten (#1359)
- Tighten `isNonNegNumber` with `Number.isFinite` so `Infinity` and
`NaN` are rejected at the read boundary, matching the write-side
intent (finding D). Hand-edited `connectionTimeout: Infinity` would
have passed the prior check and propagated to the form.
- Add a one-line comment in `normalizeMcpServers` noting that the
per-entry object is mutated in place via the `delete` calls — safe
for today's callers (freshly parsed JSON, no other reference held)
but worth flagging for the next reader (finding A).
Skipping findings B (move type guards to module scope — single call
site) and C (`isOauthObject` lenient on extra keys — reviewer
confirmed this is the intended forward-compat behavior, matching
`normalizeServerType`'s unknown-key passthrough).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* docs(servers): fix mutation-comment caller branch label (#1359)
The previous comment named the GET handler's "seed branch" as a caller
of normalizeMcpServers, but that branch returns DEFAULT_SEED_CONFIG
directly without normalizing. The actual second caller is the
file-present branch (after parseStore). Correcting the label so a
future reader looking for the seed branch caller doesn't waste time.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat(servers): hot-reload server list on external mcp.json edits (#1345)
Watch ~/.mcp-inspector/mcp.json from the backend and push a change event
on a new /api/servers/events SSE channel. The web client subscribes via
fetch + ReadableStream (so the existing bearer-auth contract is
preserved — EventSource can't send custom headers) and re-fetches on
each event. The watcher is lazy: it starts on the first SSE subscriber
and stops when the last one disconnects, so tests that never open the
channel don't spin up a real fs watcher. Self-fires from the backend's
own POST/PUT/DELETE are suppressed by capturing the post-write mtime and
comparing it on each watcher event.
Chokidar is used for cross-platform editor-save semantics (temp-file +
rename produces unlink+add sequences that bare fs.watch handles poorly
on macOS/Linux); awaitWriteFinish coalesces those into a single event.
createRemoteApp now returns a close() to release the watcher, which the
standalone server and the vite dev plugin chain into their existing
close paths.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(servers): address PR review on hot-reload (#1357)
- writeMcpAndTrackMtime now stat()s the file before writing; if the
mtime differs from the last value we recorded, an external editor
slipped in between our writes and chokidar would coalesce both into
one event whose mtime matches our own write — broadcast directly so
peer subscribers learn about the external edit (review finding 1, 5).
- Split useServers.refresh into a public refresh() (loading toggle) and
an internal refreshInternal(background) the SSE handler calls; the
list now stays visible during background refreshes instead of
flashing skeleton state on every external mcp.json edit (finding 2).
- Coalesce multiple \n\n frames landing in one decode chunk to a single
refresh call so back-to-back broadcasts don't race two concurrent
GETs whose setState order is unspecified (finding 3).
- Clarify createRemoteApp's close() docblock about in-flight SSE stream
callbacks (finding 7).
- Add integration coverage for PUT/DELETE self-suppression and for the
external-edit-between-backend-writes broadcast (finding 6).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(servers): address pass-2 review on hot-reload (#1357)
- Clarify writeMcpAndTrackMtime comment so a future reader can't
mistake the broadcast for a content-preservation guarantee: the
external edit's content may already have been read-around and
overwritten depending on the interleave with readMcpConfig.
- Drop the no-op GET in the PUT/DELETE self-suppression tests. The
pre-stat branch is short-circuited by `lastWrittenMtimeMs === null`
with or without it; matches the POST test's structure.
- Add explicit no-flicker regression test: tracks every `loading`
value across renders during an SSE-driven refresh and asserts none
flipped back to true. Catches a future regression that reaches for
setLoading(true) in the background path even under React batching.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat(servers): persist per-server settings to mcp.json (#1352)
- Add optional `settings` node alongside the transport config in each
`~/.mcp-inspector/mcp.json` entry. Round-trips through
`mcpConfigToServerEntries` / `serverEntriesToMcpConfig` and the
`/api/servers` POST/PUT routes.
- Drop `MCPServerConfig.headers` from `SseServerConfig` /
`StreamableHttpServerConfig` and remove the headers textarea from
`ServerConfigModal`. Custom headers are now entered only in
`ServerSettingsForm` and reach the wire via `settings.headers`.
- Wire `ServerSettingsModal` into `App.tsx`; saves dispatch through a new
`useServers.updateServerSettings` mutator. `updateServer` carries the
existing settings across config-only saves so they aren't dropped.
- Apply settings on the *first* outbound request after a server loads —
no need to open the settings form first:
- `settings.headers` → SSE / streamable-http wire headers.
- `settings.metadata` → `InspectorClientOptions.defaultMetadata`,
merged into every outgoing `_meta` (call-time keys win).
- `settings.requestTimeout` → `InspectorClientOptions.timeout`.
- `settings.connectionTimeout` → Promise.race wrapper around connect().
- `settings.oauth*` → pre-seeded OAuth client credentials.
- `RemoteConnectRequest` carries `settings` so the remote backend can
source transport headers from the same source as the local path.
- `oauthClientSecret` is persisted in `mcp.json` alongside stdio `env`
values, both protected by the file's `0o600` permission.
- Update `specification/v2_servers_file.md` with the new shape and the
removal of the legacy headers field.
Tests: round-trip in `serverList.test.ts`; persistence on POST/PUT in
`servers-route.test.ts`; first-connect header injection in
`transport.test.ts`; `defaultMetadata` merging in
`inspectorClient.test.ts`; `useServers.updateServerSettings` +
`updateServer` settings-preservation in `useServers.test.tsx`.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(servers): rebuild InspectorClient on every connect so settings refresh
The previous onToggleConnection branch reused the existing InspectorClient
on a same-server reconnect, which froze settings at the moment the client
was first constructed. Editing headers, metadata, timeouts, or OAuth
credentials in the settings form was therefore invisible until the user
either switched servers or reloaded the page — surprising right after a
visible save.
Always call setupClientForServer on (re)connect so the latest
target.settings flow through to the transport's headers / requestInit, to
defaultMetadata, and to the OAuth manager.
Also adds remoteClientTransport tests that pin the wire-level contract:
settings round-trip through the /api/mcp/connect body when supplied, and
the field is omitted when no settings are configured.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(history): stop wiping pendingRequestEntries inside the connect event
State managers (ManagedToolsState, ManagedPromptsState, ManagedResourcesState,
ManagedResourceTemplatesState, ManagedRequestorTasksState) handle the
client's "connect" event by calling `void this.refresh()`. The synchronous
portion of refresh runs through SDK Protocol.request → transport.send →
MessageTrackingTransport.send → trackRequest, all before yielding to the
network — so each list request gets tracked into MessageLogState's
pendingRequestEntries inside the same connect event dispatch.
MessageLogState's own "connect" listener was registered after those state
managers (it's the last one constructed in App.tsx) and cleared
pendingRequestEntries on every connect. Listener order meant the clear ran
*after* the requests were tracked but before their responses arrived,
leaving the history with unmatched "response" entries marked PENDING.
The clear-on-disconnect (via statusChange → disconnected) already handles
session-boundary cleanup, so clear-on-connect was purely defensive and now
actively harmful. Drop the connect listener; keep the disconnect path.
Tests: updated the unit test to assert the new behavior (connect does not
clear); added an integration regression that wires a sibling connect
listener firing listTools and asserts the response folds into the request
entry instead of appearing as an orphan.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* chore: drop accidentally tracked debugging screenshot
pending-requests.png was a local debugging artifact attached to a
conversation, not source. Remove it from the tree (the working copy is
kept so the user still has it locally) and rely on the existing untracked
behavior going forward.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* test(servers): add e2e check that settings.headers reach upstream MCP
Stitches the full v2 pipeline together in one assertion:
browser-style RemoteClientTransport
→ POST /api/mcp/connect (settings in body)
→ backend createTransportNode applies settings.headers to requestInit
→ SDK StreamableHTTPClientTransport
→ upstream MCP test server
Captures the upstream HTTP request via FetchRequestLogState and asserts
both `X-Tenant` and `X-Trace` are present on the outbound POST. Catches
any regression that breaks the wire between the form and the upstream
server in a single integration test, instead of the previous coverage
which exercised each hop in isolation.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(ServerSettingsModal): invert ListToggle compact prop so icon matches state
ListToggle's `compact` prop represents the list's *current* compact state:
true means "currently compact" and renders the expand icon to advertise
the action (open up). The settings modal was passing
`compact={allExpanded}` — exactly inverted. When all sections were open
it showed the expand icon (suggesting "open more") and when collapsed it
showed the collapse icon, opposite of every other ListToggle caller in
the app.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* style(HistoryListPanel): match ServerListScreen Export button — rename + default variant
Renames "Export JSON" to "Export" and switches from the subtle ToolbarButton
to the same `Button variant="default"` used by ServerListControls so the
two Export affordances look identical across screens. Also moves the
button to the right of the ListToggle (mirroring the order on the server
list screen) and disables it when there are no entries to export.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* style(HistoryListPanel): move Clear button to top toolbar, match Export style
Moves the Clear affordance out of the "History (N)" section header and
up to the panel toolbar alongside Export, where it now uses the same
`Button variant="default"` styling. Order in the toolbar is now:
ListToggle · Clear · Export. Clear is disabled when there are no
unpinned entries to remove (pinned entries are intentionally preserved).
ToolbarButton helper is unused after this change and dropped.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* chore: ignore local debugging screenshots at repo root
Stops `git add -A` from re-staging the `*.png` files Claude Code drops
at the repo root from attached conversation screenshots. The pattern
only matches the top-level (`/*.png`) so legitimate image assets in
subdirectories (e.g. `clients/web/public/`) still get tracked when
explicitly added.
Also untracks the two screenshots that slipped through previously
(initialize-request.png, pending-requests.png).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* style(LogStreamPanel): match HistoryListPanel button styling
Clear / Export / Copy All on the Logging screen were rendered with the
subtle ToolbarButton helper while History's equivalents use
`Button variant="default"`. Switch all three to the default variant so
the toolbars look identical across screens and mirror the disabling
behavior — each button is now disabled when `entries.length === 0`.
ToolbarButton helper is unused after this change and dropped.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(servers): address PR review — debounce save, preserve-on-omit PUT, validate shape, lift connect timeout
Round-up of fixes from https://github.com/modelcontextprotocol/inspector/pull/1353#issuecomment-4526861959:
1. ServerSettingsModal saves were firing on every keystroke (one PUT +
full mcp.json refresh per character). App.tsx now debounces by 300ms
and flushes on modal close so a burst of edits coalesces into one PUT
and the form value can't flicker back mid-typing.
2. `PUT /api/servers/:id` with an omitted `settings` field used to
silently wipe the persisted settings node. The route now treats the
three intents distinctly:
- field omitted → preserve existing on-disk settings
- explicit null → clear the settings node
- settings object → validate + apply
`useServers.updateServer` is simplified to drop the client-side
carry-through hack, and the test that documented the broken behavior
is flipped + augmented with explicit-clear and bad-shape cases.
3. `connectionTimeout` Promise.race is moved from App.tsx into
`InspectorClient.connect()`, reading `this.serverSettings?.connectionTimeout`.
TUI/CLI consumers now get the same behavior; on timeout the client
internally disconnects so the next connect starts clean.
4. Custom Headers hint in `ServerSettingsForm` notes that `Authorization`
is owned by the OAuth flow when OAuth is configured — avoids the
sharp edge where the SDK's authProvider silently overwrites a
user-set Authorization header.
5. New `validateSettings` on the server route rejects malformed shapes
(`settings: []`, `settings.headers: "oops"`, non-numeric timeouts,
etc.) with 400 instead of letting them persist and crash the UI.
7 (nit). Renamed `settings` → `savedSettings` inside `setupClientForServer`
so the variable holding the persisted settings doesn't shadow the
later locally-derived values.
Follow-ups opened separately:
#1356 — Move oauthClientSecret out of mcp.json into the OS keychain (review #6).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(servers): address PR review pass 2 — connect-timeout test, unmount cleanup, tagged intent, lenient settings read
Pass-2 review at https://github.com/modelcontextprotocol/inspector/pull/1353#issuecomment-4527089705:
(a) New integration test pins the new client-layer connect-timeout
contract: a hanging stub transport with `connectionTimeout: 50`
rejects with the descriptive error and lands the client in "error"
status. Documents what TUI/CLI consumers should expect from the
lifted Promise.race in InspectorClient.connect().
(b) useEffect cleanup clears `pendingSettingsTimerRef` on unmount so a
route change or HMR mid-debounce can't fire one stale PUT against
an unmounted component.
(c) Refactor the PUT-route `settingsIntent` discriminator from
`"preserve" | "clear" | { value }` to a tagged union with `kind`,
and consume it with a `switch` for exhaustiveness — reads cleaner
and makes adding a fourth intent painless.
(d) `normalizeMcpServers` now runs `validateSettings` on the on-disk
settings node and silently strips it if invalid, mirroring the
existing lenient-on-read pattern from `normalizeServerType`. Closes
the read/write asymmetry where a hand-edited malformed `settings`
could survive on disk and get carried through preserve-on-PUT.
New route test pins the strip-on-GET + no-propagate-on-preserve-PUT
behavior.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(servers): pass-3 review — patch-style PUT, toast on close-flush fail
Pass-3 review at https://github.com/modelcontextprotocol/inspector/pull/1353#issuecomment-4529276897.
(b) `PUT /api/servers/:id` now treats both `config` and `settings` as
optional patches:
- field omitted → preserve the on-disk value
- explicit body value → apply
- settings: null → clear
`updateServerSettings` stops reading `existing.config` from the
in-memory `servers` snapshot, which closes the closure-capture
hazard where a refresh between debounce-schedule and flush could
silently revert a separate edit. The route is the single source of
truth for transport config now. New tests cover settings-only PUT,
rejected non-object config, and the empty-body no-op patch case.
The previous "rejects a missing config" test is rewritten to
document the new no-op-preserve semantics.
(c) Inline comment near the PUT rebuild loop documents that writing
the full mcpServers map back is what gives the route its desirable
self-healing property for malformed settings on *other* entries —
deliberate side-effect, not an accident of normalization.
(d) `flushPendingSettings` surfaces failures via @mantine/notifications
instead of swallowing them. The modal already closed by the time
the flush PUT resolves, so a silent fail would leave the user
thinking their last edits saved (especially painful for the OAuth
client secret). The notifications wiring was already mounted in
main.tsx.
Pass-3 (a) — settings form being effectively uncontrolled mid-edit — is
acknowledged in the reply: working as intended today; the proper fix is
internal modal state when file-watching (#1345) forces the issue.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(servers): pass-4 review — strip smuggled settings out of config
Pass-4 review at https://github.com/modelcontextprotocol/inspector/pull/1353#issuecomment-4529861349.
Main finding: `normalizeServerType` spreads unknown keys, so a body
that nested `settings` inside `config` would smuggle a settings field
onto the stored entry without ever passing through `validateSettings`.
Strip the settings key off the incoming config in `buildStoredEntry`
so `validateSettings` remains the single write path for the settings
node — the "one source of truth" invariant pass 2 set up holds again.
Two new tests pin the strip on both routes:
- POST with `config.settings = { bogus }` lands an entry with no
settings node on disk.
- PUT with `config.settings = { bogus }` + `settings: null` clears
the real settings *and* doesn't re-attach the bogus payload via
the spread.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(servers): pass-5 review — absorb late connect rejection, tighten validateSettings, log smuggle warning
Pass-5 review at https://github.com/modelcontextprotocol/inspector/pull/1353#issuecomment-4529954490.
Main finding: when the connect-time `Promise.race` lost to the timeout,
the original `this.client.connect(this.transport)` promise was left
pending. After `disconnect()` tore the transport down, real transports
(SSE / streamable-http) would reject that promise later — with no
handler attached, producing a Node `unhandledRejection` (and a vitest
suite warning). Attach a no-op `.catch` right after creating the
connect promise so Node sees the late rejection as handled.
Nit (a): `validateSettings` was casting the raw object through to
`InspectorServerSettings`, which let unknown keys ride along onto disk.
Replaced the cast with an explicit pick-and-build over the known
fields. Unknown stowaways now silently drop on the way through. New
route test pins the behavior.
Nit (b): `buildStoredEntry` now logs a `warn` via the route's pino
logger when an incoming `config` carries a `settings` key. The strip
is still defensive (it isn't a documented contract) but the warning
lets a misconfigured client find out about the wasted field instead
of debugging a silent drop.
Nit (c) — the defensive `?? {}` fallback in `buildStoredEntry` —
acknowledged in the review reply; left in place as belt-and-braces.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(servers): pass-6 polish — scope absorber, thread id, drop empty OAuth, guard `in`
Pass-6 review at https://github.com/modelcontextprotocol/inspector/pull/1353#issuecomment-4529979135. Reviewer signed off as ready; landing the four marginal cleanups since they're all small.
(a) `connectPromise.catch(() => {})` moves inside the
`if (connectTimeoutMs > 0)` branch. The await-only path propagates
rejection cleanly through the outer try/catch — the absorber is
only needed when the race leaves the original promise unattended.
Code now reads as "absorber is tied to the race," which is the
real invariant.
(b) Thread `id` (POST id, PUT newId) through `buildStoredEntry` and
include it in the smuggle warning's pino bindings. Two clients
smuggling simultaneously are distinguishable in the log instead
of indistinguishable.
(c) `validateSettings` coerces empty-string OAuth fields to absent.
The form emits `""` when the user clears an input, and an empty
`oauthClientId` on disk would later be misread as "OAuth configured."
New route test pins it.
(d) `buildStoredEntry` adds a null/object guard before `"settings" in
configObj`. The route layer pre-checks this on POST and PUT, but
the helper's `unknown` parameter is a weak contract; the guard
makes it safe to call from anywhere.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* docs(spec): reflect settings persistence + patch-PUT contract in v2_servers_file
The doc was authored when `mcp.json` only stored the server list. Bring it
forward to match what shipped in #1352 / PR #1353:
- Summary calls out that the file also stores per-server settings and
links into the Per-server settings section.
- On-disk format note explains that `settings` is an optional Inspector-
specific extension that other MCP tools ignore.
- New top paragraph on the Per-server settings section explains the
basic-config / settings UI split rationale (simpler form for the
common case, advanced settings tucked behind a second dialog).
- UI bullet is rewritten to match the patch-style PUT semantics that
emerged across the review cycle: settings-only and config-only
PUTs preserve their counterpart inside the write lock, so neither
modal can silently clobber the other half. The earlier "must
re-send settings on PUT" hack is gone.
- New bullets capture the save cadence (300ms debounce + flush-on-close
+ failure toast), the patch-PUT contract (omit-preserve / null-clear /
rename), and the write-path / read-path validator gates.
- Secret storage bullet links to #1356 (OS-keychain follow-up).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* docs(spec): show a third entry with the optional settings node
The On-disk format example demonstrated only basic stdio entries. Add a
streamable-http entry with a populated `settings` node so readers can
see the shape inline rather than having to scroll to the Per-server
settings section. Inline comment in the block cross-links there for the
full contract and notes that other MCP tools ignore the extension key.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat(servers): export server list as mcp.json download (#1346)
Adds an Export button to the Servers screen, next to "Add Servers".
Clicking it downloads the current server list as a canonical mcp.json
file (2-space indent, same shape serializeStore writes), so the user
can hand it off, sync it to another machine, or symlink into Claude
Desktop / Cursor / Cline.
Wiring:
- ServerListControls: new Export Button (variant=default), disabled
when serverCount === 0 (nothing to download).
- ServerListScreen / InspectorView: thread `onExport` / `onServerExport`
through.
- App.tsx: onServerExport runs serverEntriesToMcpConfig on the in-memory
servers list, JSON.stringify with 2-space indent, and triggers a Blob
download via a temporary anchor (appended to body for Firefox
compatibility, then removed; object URL revoked).
Tests:
- ServerListControls unit tests cover button-count (toggle hidden vs
shown, Export always present), disabled-on-empty, click fires onExport.
- ServerListControls.stories grows play functions for both stories —
WithServers asserts Export is enabled + click fires the spy in real
Chromium, WithoutServers asserts Export is disabled.
- Existing screen + view tests / stories updated for the new prop.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(servers): address PR #1351 review (extract helpers + exception safety + guard)
Substantive items from the review:
1. Serialization format parity (review #1): added
`serializeMcpConfig(entries)` to core/mcp/serverList.ts — a
browser-safe helper that runs `serverEntriesToMcpConfig` then
stringifies with 2-space indent. Single source of truth for export
formatting; can't drift from serializeStore's on-disk shape.
2. Exception safety on the anchor cleanup (review #2) + a direct test
for the App.tsx download wiring (review #6): extracted
`downloadJsonFile(filename, json)` to clients/web/src/lib/
downloadFile.ts. removeChild + revokeObjectURL run in a `finally`
so a thrown click() doesn't leak the DOM node or the object URL.
Unit-tested under happy-dom with URL.createObjectURL mocked
(happy-dom doesn't ship it).
3. Empty-list guard (review #5): App.tsx onServerExport bails when
`servers.length === 0`. The button's disabled prop already prevents
the UI path, but the handler is now locally correct against any
future programmatic caller (storybook, keyboard shortcut, etc.).
Acknowledged but not changed:
- Tooltip on the disabled Export button (review #3) — Mantine's
disabled-button + Tooltip wart wants a span wrapper + custom events
config. Optional per the reviewer; deferred.
- Repeated exports → `mcp (1).json` browser suffix (review #4) — a
deliberate trade-off since `mcp.json` is the conventional name a user
would expect to symlink into Claude Desktop / Cursor / etc.
Tests: 1647 unit (+6) + 410 integration + 306 storybook = 2363 green.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat(servers): persist server list to ~/.mcp-inspector/mcp.json (#1343)
Replaces the hardcoded SEED_SERVERS in App.tsx with a file-backed list at
~/.mcp-inspector/mcp.json. Full CRUD round-trips through new /api/servers
routes; first launch writes the two existing seeds if the file is absent.
Backend (core):
- store-io: getDefaultMcpConfigPath() — sibling of getDefaultStorageDir()
- core/mcp/serverList.ts: pure converters MCPConfig <-> ServerEntry[],
DEFAULT_SEED_CONFIG, and normalizeServerType (moved from node/config.ts
to keep serverList.ts Node-free)
- core/mcp/remote/node/server.ts: GET/POST/PUT/:id/DELETE/:id /api/servers
routes, plus RemoteServerOptions.mcpConfigPath defaulting to
getDefaultMcpConfigPath(). Ids validated with validateStoreId.
Frontend:
- core/react/useServers.ts: { servers, loading, error, refresh, addServer,
updateServer, removeServer } over the new endpoints
- App.tsx: drops SEED_SERVERS, wires useServers, disconnects on remove of
the active server, follows id on rename of the active server
- ServerConfigModal: Add/Edit/Clone form (stdio: command/args/env/cwd;
sse/streamable-http: url/headers as one-per-line text)
- ServerRemoveConfirmModal: confirm-on-remove with id + transport summary
Tests (52 new):
- unit: getDefaultMcpConfigPath env-var permutations, serverList round-trip,
hook flows against real Hono app.fetch (no TCP), modal forms + validation
- integration: /api/servers routes against a real createRemoteApp + tmp file
- stories with play functions for both new modals
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(servers): address PR #1349 review (correctness + nits)
Bugs:
- normalizeServerType: validate against the allowed set and default
unknown / non-string types to "stdio" (matches the missing-type case).
Prevents a hand-edited {"type":"websocket"} from leaking past `as
ServerType` and surprising downstream narrowing.
- onConfirmRemove: when removing the active server, also destroy() the 9
state managers and null them + the client out so they GC immediately
instead of waiting for the next server switch.
- /api/servers mutating routes: serialize POST/PUT/DELETE through an
in-process write-lock so concurrent read-modify-write cycles can't lose
updates. New integration test fires 25 parallel POSTs and asserts every
entry lands.
- readMcpConfig comment: update to match actual behavior (valid-JSON file
with no `mcpServers` → empty; invalid JSON → 500 from the outer try, so
corruption surfaces rather than silently emptying the list). New
integration tests pin the 500 path on both GET and POST.
Nits:
- Add --inspector-surface-subtle token, use it in ServerRemoveConfirmModal
instead of the raw Mantine var.
- Rename useServers buildHeaders boolean param to includeJsonBody.
- Add a comment on the env-value vs header-value trim asymmetry (env
values preserve trailing whitespace; header values follow RFC 7230).
Tests: 1630 unit + 404 integration + 306 storybook all green.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(ServerConfigModal): null currentTarget in functional setState updater
React nulls SyntheticEvent.currentTarget after the synchronous handler
returns. Reading e.currentTarget.value inside `setForm((f) => ...)`
worked under happy-dom (which keeps currentTarget around) but threw
"Cannot read properties of null (reading 'value')" on the first
keystroke in a real browser. Capture the value before the functional
updater closes over it.
All seven onChange handlers in ServerConfigModal had the bug. Extended
the AddEmpty storybook play to type into Server ID and Command — the
real-Chromium runner now exercises this path so a future regression
fails the test suite instead of the user.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(remote-server): surface stderr + transport_error when subprocess crashes during startup
A stdio subprocess that spawns successfully but exits before the browser
opens the SSE events stream used to vanish into a bare "Session not
found" 404. The fix has three parts:
1. RemoteSession.markTransportDead always pushes the transport_error
event (via pushEvent's queue) rather than dropping it when no
consumer is attached. The queued event drains when the consumer
eventually arrives.
2. /api/mcp/connect's transport.onclose handler no longer deletes the
session immediately when no consumer is attached. Instead it holds
the session (with queued stderr + transport_error) for a 30s grace
window so the events endpoint can drain it. A timeout sweeps if no
consumer ever connects.
3. /api/mcp/events detects isTransportDead after attaching the consumer
(which just drained the queue), yields once to flush, then closes
the stream and deletes the session.
End result: the user sees the actual error (e.g. "Cannot find module"
from stderr + the transport-closed reason) instead of a confusing 404.
Tests:
- New unit tests (clients/web/src/test/integration/mcp/remote/remote-session.test.ts)
cover the queueing behavior, live delivery, and clearEventConsumer
signaling under all transport-state permutations.
- New integration test (connect-crash.test.ts) spawns a node subprocess
that writes to stderr and exits, then verifies the events stream
delivers both the stderr line and the transport_error event before
closing — the exact scenario from PR #1349.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(state): gate listRequestorTasks on the server's `tasks` capability
ManagedRequestorTasksState was calling tasks/list on every connect.
Servers that don't advertise the `tasks` capability (the common case)
reply with -32601 "Method not found", which then surfaces in the
inspector's console for every connect.
Skip the call when the server doesn't declare `tasks` — empty tasks
list is the right semantics. Existing tests opt into capabilities so
the live-path coverage is preserved; two new tests pin the gated
behavior (direct refresh + connect-event-triggered refresh).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(App): clear activeServerId on disconnect so other server cards re-enable
ServerCard dims any card whose id differs from `activeServer` (and marks
it inert), so cards stay non-interactive after a disconnect until the
user reconnects elsewhere. activeServerId was set on connect but never
reset on disconnect, leaving the whole screen locked to the just-
disconnected card.
Subscribe to InspectorClient's `disconnect` event and clear
activeServerId there. Covers all three disconnect paths uniformly:
- user toggles the connected card off
- user clicks the header Disconnect button
- mid-session transport failure (subprocess exit / network drop)
Event-based instead of effect-based on connectionStatus to avoid the
first-render-clobbers-new-id trap (status starts as "disconnected" for
a freshly-created client before connect() runs).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Design for replacing the hardcoded SEED_SERVERS in App.tsx with a
file-backed list at ~/.mcp-inspector/mcp.json, reusing the v1.5-ported
storage primitives and the canonical mcp.json parser. Ties the new doc
into the Storage section nav as the first sub-menu item.
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* Port resource preview UX improvements to PromptsScreen (#1328)
Brings the prompt-fetching flow up to parity with the resource preview
work from #1326:
- Cap the argument-form pane at maw=40% so a bare input + button doesn't
stretch across wide displays.
- Auto-fetch no-argument prompts the moment they're picked from the
sidebar (handleSelectPrompt routes them straight to onGetPrompt({})),
and hide the form once the user clicks Get Prompt — the result panel
takes the same fixed-height column the resource preview uses.
- Apply the PreviewCard variant=preview pattern: card sizes to content
but caps at viewport, PromptMessagesDisplay pins its header (flex 0 0
auto) and lets the inner ScrollArea (flex 0 1 auto, mih 0) absorb
overflow.
- Tag getPromptState with the prompt name in App.tsx so the screen can
ignore a stale result whose name no longer matches the selection.
- MessageBubble now routes each content block through ContentViewer
(markdown for text via mimeType="text/markdown", image/audio/resource
via existing ContentViewer branches). Non-renderable block types
(tool_use, tool_result) are filtered out; the role-label header keeps
the bubble visible regardless.
- PromptArgumentsForm gains onCompleteArgument + completionsSupported;
when both are set, each input becomes Mantine Autocomplete with the
same per-arg AbortController + per-arg debounce timer pattern as
ResourceTemplatePanel. PromptsScreen + InspectorView + App.tsx wire
the callback to inspectorClient.getCompletions with a
ref/prompt envelope.
Closes#1328.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* Add close (X) buttons to ResourcePreviewPanel and PromptMessagesDisplay
A top-left CloseButton dismisses the preview panel. The host screen
decides what to fall back to:
- ResourcesScreen remembers the originating template URI when the user
reads from a template form (originatingTemplateUri); closing the
preview restores the template form. Plain-resource reads (sidebar
selection) just empty the selection and return to the empty state.
Picking a different template / resource from the sidebar clears the
back-trail.
- PromptsScreen flips submittedFor back to undefined when the closed
prompt has arguments — the argument form re-renders with the
user's values preserved so they can edit and re-submit. No-arg
prompts have nothing to fall back to, so the selection is dropped
and the empty state appears.
The X button is hidden when the host omits onClose, so callers that
don't want a dismiss control keep their existing rendering.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* Add close button to pending and error preview states
The OK-state preview already had its X button (inside the panel
component itself). Pending and error states were inline cards in
renderReadState / renderPreview with no way to dismiss them, so a
user who submitted bad input to a resource template or prompt was
stuck staring at the error until they picked a different sidebar
item. Adds a top-left CloseButton row to both states on both screens,
wired to the same handleClosePreview handler — so the template form
or argument form re-appears on close just like it does from the OK
state.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* Fire prompt completions on focus and send every sibling in context
Two fixes to the PromptArgumentsForm autocomplete:
- Focusing an argument input now fires completion/complete immediately
(handleFocus) — the dropdown is populated as soon as the user clicks
in, not only after they start typing. Any in-flight debounce timer
for the same arg is cancelled so a stale keystroke request can't
overwrite the fresh focus response.
- The completion context now includes every declared prompt argument
(with "" for ones the user hasn't typed into yet), minus the one
being completed. Previously the context only carried args the user
had touched, so servers that disambiguate based on co-arguments
couldn't see the full picture on the first keystroke.
Tests cover both: a focus-only path that fires before any keystroke,
and a typing path that asserts the empty sibling is sent through.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* Fire resource-template completions on focus
Mirrors the prompt-side change: focusing a template variable input
now fires completion/complete immediately so the dropdown populates
the moment the user clicks in, rather than waiting for the first
keystroke + 300ms debounce. Any pending debounce timer for the same
variable is cancelled first so a stale keystroke response can't
overwrite the focus response.
The sibling-context coverage was already correct here — `variables`
is seeded with empty strings for every declared template variable at
mount and on template switch, so the context payload always carries
the full variable set minus the one being completed.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* Style Autocomplete dropdowns to match the rest of the app
Two fixes so the completion dropdowns on prompt-argument and
resource-template inputs look like every other dropdown in the app:
- App.css: the dark-mode dropdown / option-hover rules used to target
.mantine-Select-* only, so the Autocomplete dropdowns shipped with
Mantine's default surface color and a different hover background.
Re-scoped to .mantine-Combobox-* so every Combobox-built input
(Select, Autocomplete, MultiSelect, …) inherits the same styling.
- theme/Autocomplete.ts: new theme module setting `radius: "md"` so
Autocomplete matches the default Select / TextInput corner radius.
Wired through theme/index.ts and theme/theme.ts.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* Fix dropdown styling — Select and Autocomplete need separate selectors
The previous commit re-scoped the dark-mode dropdown rules from
`.mantine-Select-*` to `.mantine-Combobox-*` on the assumption that
both Select and Autocomplete share a Combobox class on the dropdown
root. They don't — Mantine's Combobox factory takes
`__staticSelector` from the parent component, so Select emits
`mantine-Select-dropdown` and Autocomplete emits
`mantine-Autocomplete-dropdown` with no shared `Combobox` class on
that element. The `.mantine-Combobox-*` selector matched nothing,
so the TaskControls filter (and every other Select) lost its gray
surface and reverted to Mantine's default brown-ish dark background.
List both selectors so Select dropdowns keep their styling and
Autocomplete dropdowns pick it up.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* Address review feedback: canSubmit, fire-time context, re-click guard
- PromptArgumentsForm: add a canSubmit guard that disables Get Prompt
until every required argument has a value, matching the resource
template form's symmetry. Optional args may stay blank.
- Both forms: capture sibling context inside the debounce callback,
not at schedule time. Hold the latest argumentValues / variables
in a ref + sync via useEffect, and call buildContext() at fire
time. Without this, typing in arg A then arg B within the 300ms
window would ship A's request with B's pre-keystroke value.
- PromptsScreen.handleSelectPrompt: early-return when the user
re-clicks the already-selected prompt — sidebar is for navigation,
✕ is for dismiss. Previously a re-click wiped form values and
re-fired the auto-fetch for no-arg prompts.
- MessageBubble: expand the comment on effectiveMimeForBlock to flag
the unconditional markdown promotion as a known asymmetry with
the resource side (which only promotes when the server signals it),
pending a per-block mimeType in the spec.
- New tests: canSubmit disabled / enabled transitions, fire-time
context capture across siblings, abort-path verification (a stale
in-flight response must not overwrite the fresh one), and the
no-op re-click guard on PromptsScreen.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* Address follow-up review nits: ghost suggestions, dead previewActive branch
- Clear `completions[name]` at the top of handleChange in both
PromptArgumentsForm and ResourceTemplatePanel so the dropdown
doesn't show stale options from the previous prefix during the
300ms debounce + network window.
- Simplify previewActive in PromptsScreen to drop the unreachable
`!getPromptState?.promptName` fallback. App.tsx tags every prompt
state transition with `promptName`, so the fallback was dead code
that obscured the actual runtime invariant.
- Add a focused test for each form that asserts stale dropdown
options vanish the instant a new keystroke arrives.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* Expose live resource subscriptions via a hook (#1325)
Add ResourceSubscriptionsState that mirrors InspectorClient's subscribed
URIs as InspectorResourceSubscription[], resolving each URI against the
managed resources list (so the subscription tile shows server-supplied
name/title) and stamping lastUpdated on notifications/resources/updated.
Pair it with useResourceSubscriptions and wire App.tsx so the Resources
screen reflects subscribe/unsubscribe actions in real time.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* Address review notes on resourceSubscriptionsState
- Soften the onResourceUpdated comment: the client's dispatch is already
guarded by subscribedResources.has(uri), so the re-check is true
defense-in-depth rather than guarding a known hazard.
- Use this.getSubscriptions() in the statusChange handler so every emit
goes through the defensive-copy path.
- Document the deliberate fallback to a synthetic Resource when a
previously-listed resource is removed from the managed list while
the user is still subscribed.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* Reflect subscribed state in preview + compact subscribed tile labels
- Wire `isSubscribed` on the ReadResourceState passed to ResourcesScreen
by deriving it from the live subscriptions list in App.tsx. The
ResourcePreviewPanel's SubscribeButton already flips its label to
"Unsubscribe" when subscribed; without this derivation isSubscribed
was always false and the button looked stuck on "Subscribe".
- In ResourceSubscribedItem, display only the last URI path segment
(truncated with ellipsis if it still overflows) and surface the full
URI via a hover tooltip. Keeps the Subscriptions pleat readable when
resources have long path-style URIs.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* Auto-read resource on sidebar click
Clicking a resource in the URIs accordion now triggers onReadResource
in addition to setting the selection, so the preview panel jumps
straight to the pending → ok render path. Removes the unreachable
"Click to read this resource" placeholder, since the loader / preview /
error states always cover the rendered output for a selected resource.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* Render markdown resources via react-markdown + fix preview overflow
- Add react-markdown / remark-gfm dependencies.
- ContentViewer accepts an optional mimeType prop; when it matches
text/markdown (or text/x-markdown), text blocks render via
react-markdown wrapped in a `.markdown-content` container. Non-markdown
text now uses the existing Code "wrapping" variant so long lines stay
inside the panel.
- ResourcePreviewPanel infers `text/markdown` from a `.md` / `.markdown`
URI suffix when the server didn't supply a mimeType, and threads the
effective mime through to ContentViewer per content item.
- ResourcesScreen flex children get miw=0 so a long unwrappable line in
the resource body can no longer push the preview past the viewport's
right edge.
- App.css gains a `.markdown-content` ruleset constraining nested pre,
table, code, and img elements to the container's width.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* Contain resource preview: sticky header/footer, scrollable body
Restructure ResourcePreviewPanel into a fixed-height column with the
resource title + URI pinned to the top and the timestamp / annotations /
subscribe-refresh actions pinned to the bottom. The content viewer area
in the middle now owns its own ScrollArea, so a long markdown body
scrolls within the panel instead of pushing the subscribe button below
the viewport.
In ResourcesScreen, the selectedResource branch (and the template
branch's right pane) now hosts the panel inside a PreviewPane Flex
column with the screen's max-height, and renderReadState wraps in a
FillDetailCard sized to fill that column. The legacy outer
ScrollArea.Autosize is removed for these panes since scrolling is
internal now.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* Auto-hide template after read; size preview to content with cap
- After the user clicks Read Resource on the template form,
handleReadResource now clears selectedTemplateUri so the screen
swaps from the template form to the resource preview. Previously
both panels stayed mounted side-by-side.
- The preview no longer hard-fills the viewport. The Card uses a new
"preview" theme variant (overflow: hidden) and is content-sized,
capped at SCROLL_MAX_HEIGHT. ResourcePreviewPanel's flex column
marks the header / meta / footer rows as `flex: 0 0 auto` and the
ContentScroll as `flex: 0 1 auto` with `mih: 0`, so:
- short content → card hugs it, footer sits right under the body
- long content → card caps at viewport, ContentScroll shrinks
and scrolls internally, footer stays pinned at the cap
- ScrollArea / Group imports in ResourcesScreen pruned to match the
simpler single-pane layout.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* Cap template form panel at 40% width
A bare text input + Read Resource button stretched across the whole
main content area looks awkward on wide displays. Apply maw=40% to
the PreviewPane in the template branch so the form keeps a reasonable
form-field width regardless of viewport.
The preview branch is unaffected — resource bodies still get the full
main area for content.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* Live completion/complete suggestions on resource template inputs
Wires the server's `completions` capability through to the resource
template form so each variable's input becomes an Autocomplete that
fires `completion/complete` (debounced, 300ms) on every keystroke and
surfaces the returned values as a dropdown — mirroring v1's behavior.
- core/mcp/inspectorClientProtocol.ts: surface getCompletions on the
protocol so non-runtime callers (state managers, hooks, tests) can
depend on it. FakeInspectorClient gains a vi.fn-backed stub.
- ResourceTemplatePanel: accepts onCompleteArgument + completionsSupported.
When both are present it renders Mantine Autocomplete instead of
TextInput, debounces keystrokes via per-arg timers, aborts in-flight
requests on the next keystroke, and disables client-side filtering
(the server already filtered for the typed prefix).
- ResourcesScreen / InspectorView: thread the props through; the
screen-level callback re-injects the active template's URI as the
`ref: "ref/resource"` so the panel-level callback stays ref-free.
- App.tsx: wires onCompleteArgument to inspectorClient.getCompletions
and derives completionsSupported from `capabilities?.completions`.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat(web): wire App.tsx to v2 core/ hook layer; remove demo stub from InspectorView
App.tsx mounts InspectorView with live data from the v2 hook layer instead
of the title-page placeholder. InspectorView becomes pure prop-driven; the
demo handshake stub and connection state machinery are removed.
Connection lifecycle:
- App.tsx lazily instantiates `InspectorClient` on the connect edge,
rebuilding it (and its state managers) when the user picks a different
server. State managers (Managed{Tools,Prompts,Resources,ResourceTemplates,
RequestorTasks}State + MessageLog/FetchRequestLog/StderrLog) are
destroyed on switch.
- `useInspectorClient` + the per-primitive `useManaged*` hooks drive the
view's data. `latencyMs` is captured at the connecting → connected edge
via a ref so the intermediate rerenders don't lose the start timestamp.
- `errorMessage` is captured from `client.connect()` rejection; status-
machine-driven errors that don't surface through the promise are a known
follow-up (no `error` event in the v2 InspectorClientEventMap yet).
Action handlers:
- `onCallTool` / `onGetPrompt` / `onReadResource` await the corresponding
InspectorClient method and reflect pending → ok/error in `ToolCallState`
/ `GetPromptState` / `ReadResourceState` panel props. Manually verified
end-to-end against `npx @modelcontextprotocol/server-filesystem` —
the Tools result panel renders the real "Allowed directories: /private/tmp"
output rather than a stub.
- `onSetLogLevel` calls `client.setLoggingLevel(level)` and optimistically
bumps `currentLogLevel` locally (the request has no echo notification).
- `onSubscribeResource` / `onUnsubscribeResource` / `onCancelTask` route to
their respective client methods. Server CRUD (`onServerAdd`/`onServerEdit`
/…) and history pinning are intentionally `todoNoop` for now; the seed
server list is hardcoded per the agreed scope (a `useServers` v2-only
hook is a separate effort).
- `notifications/message` notifications flowing through `MessageLogState`
are filtered into `LogEntryData[]` and passed to the Logs screen, so
`logging/setLevel` and the resulting log stream round-trip.
InspectorView (`InspectorView.tsx`):
- Removed: `STUB_*` constants, `handshakeTimer` ref, `handleToggleConnection`
/ `disconnect` functions, stub `bridgeFactory`/`sandboxPath`, internal
connection state, `logLevel` state.
- Added 38 props covering connection state, panel states, log level, and
action callbacks. The view is pure presentational.
- `availableTabs` is derived from `connectionStatus`; `activeTab` is
clamped to whatever's currently available (avoiding the `set-state-in-effect`
lint and keeping the previously-selected tab on reconnect).
Stories + tests:
- `InspectorView.stories.tsx` adds spy callbacks for every action prop and
new `Connected` / `ConnectionError` stories for the connected/error
narratives.
- `InspectorView.test.tsx` is rewritten around the prop-driven contract
(10 tests covering the empty list, server cards, toggle dispatch,
connected header, error banner, tab snap-back on disconnect, app
filtering, log-level dispatch, autoScroll local toggle).
Dev backend port wiring (`vite.config.ts`):
- Vite's dev server is now pinned to `CLIENT_PORT` (default 6274) with
`strictPort: true`, matching the dev backend's `allowedOrigins`. Without
this, Vite would fall back to 5173 while the Hono origin check still
required 6274 — breaking every browser `/api/*` fetch out of the box.
- Top-level `resolve.alias` now spreads the same bare-module aliases the
vitest projects use. Rolldown can't reach `pino/browser.js` /
`zustand/middleware` from `core/`'s parent (no node_modules there); the
wired App pulls those modules into the browser dep graph for the first
time, so the aliases need to be visible at build time too.
createWebEnvironment helper (`src/lib/environmentFactory.ts`):
- Mirrors v1.5's `lib/adapters/environmentFactory.ts`: builds
`InspectorClientEnvironment` from `createRemoteTransport` +
`createRemoteFetch` + `createRemoteLogger` + `BrowserOAuthStorage` +
`BrowserNavigation`. Returns the assembled environment plus the logger
for direct app-level use.
Closes#1244.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(web): wire dev-backend auth token into createWebEnvironment + show full stdio command in ServerCard
Two issues surfaced after the #1244 wiring landed locally:
1. **401 on `/api/mcp/connect`** when running `npm run dev` with the
default auth (no `DANGEROUSLY_OMIT_AUTH`). The dev backend generates a
random token and prints `http://localhost:6274?MCP_INSPECTOR_API_TOKEN=…`
in the banner, but App.tsx was calling `createWebEnvironment(undefined, …)`
so the browser never sent `x-mcp-remote-auth`. The Hono middleware
correctly rejected with "Authentication required. Use the
x-mcp-remote-auth header with Bearer token."
Adds a small `getAuthToken()` helper in App.tsx that reads
`MCP_INSPECTOR_API_TOKEN` off `window.location.search`, falls back to
`sessionStorage`, and persists fresh values into sessionStorage so
client-side navigation / OAuth round-trips don't drop the token. The
resolved value is passed to `createWebEnvironment` whenever a new
InspectorClient is built.
2. **Misleading ServerCard display** — the seed server's `config` is
`{ command: "npx", args: ["-y", "@modelcontextprotocol/server-filesystem",
"/tmp"] }` but the card only rendered `"npx"`, so it looked like the
command was incomplete. `getCommandOrUrl` now joins `command + args`
for stdio configs, so the card shows the same string that gets spawned.
SSE / streamable-http paths unchanged.
Manual end-to-end re-verified against `npm run dev` (no DANGEROUSLY_OMIT_AUTH):
- Banner prints the auth-token URL.
- Browser navigates to that URL → `getAuthToken()` picks the token off the
query string, persists to sessionStorage, threads it through to every
`/api/*` request.
- Server card shows `npx -y @modelcontextprotocol/server-filesystem /tmp`.
- Toggle connect → connection succeeds; ViewHeader shows
`Connected (~1.3s)`; Tools tab populates with the 14 filesystem tools.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(web): use dark-9 instead of blue-9 for dark-mode page background
`main.tsx` overrode `--mantine-color-body` to `--mantine-color-blue-9` in
dark mode, while `.storybook/preview.tsx` used `--mantine-color-dark-9`.
The dev app then rendered with a blue background while every story (and
the design intent visible in storybook) used dark grey.
Switch main.tsx to `--mantine-color-dark-9` so the dev app matches the
storybook reference.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat(web): seed the Servers screen with the "everything" reference server
Adds `@modelcontextprotocol/server-everything` to the hardcoded seed list
alongside the filesystem server. The everything server exposes tools,
prompts, resources, sampling, and completion — covering most of the
surface a developer might want to exercise against the wired-up
InspectorView without configuring anything.
Comment updated to explain the two-seed shape.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(web): address #1244 PR review feedback (JsonValue cast, client lifecycle, dead code)
- **#1 — Misleading `Record<string, never>` cast**: Imported `JsonValue`
from `@inspector/core/mcp/index.js` and changed `args as Record<string,
never>` to `args as Record<string, JsonValue>` in `onCallTool`. The new
cast honestly narrows from the screen-level `Record<string, unknown>`
to what `InspectorClient.callTool` actually requires. Added a comment
explaining the boundary.
- **#2 + #3 — Previous InspectorClient not disconnected on server switch
or unmount**: Added a `useEffect` keyed on `inspectorClient` whose
cleanup calls `inspectorClient.disconnect()`. One effect covers both
the swap-on-switch leak (prior session's transport stayed open until
GC) and the unmount-during-connected leak (HMR / tests).
`setupClientForServer`'s existing state-manager `destroy()` calls
handle the listener side; the new effect handles the transport side.
- **#4 — Dead `void` block at bottom of file**: Deleted the trailing
`void FetchRequestLogState; void StderrLogState;` block. The state
managers are already constructed via `new FetchRequestLogState(client)`
/ `new StderrLogState(client)`, so the imports are referenced by the
linter's reckoning. The `void` pair was dead code with a comment that
no longer matched reality.
- **#5 / #7 / #8 — Follow-up TODOs**: Added TODO refs to the new
follow-up issues:
- TODO(#1323): `error` event in InspectorClientEventMap so mid-session
transport failures surface (connect-only catch is incomplete).
- TODO(#1324): negotiated `protocolVersion` through `useInspectorClient`
(currently hard-coded `"2025-06-18"`).
- TODO(#1325): `useResourceSubscriptions` hook so subscribe/unsubscribe
buttons reflect their server-side effect.
Items #6 (useCallback dep churn), #9 (helper unit tests), and #10
(shared no-op bridge factory fixture) are acknowledged in the PR reply
as deferred — no functional impact and out of scope for a "wire the
hook layer" PR.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat(web): port v1.5 dev backend (vite-hono-plugin + sandbox + remote-server bootstrap)
Ports the v1.5 in-process dev backend into v2 so `npm run dev` serves the
Hono remote-server (`/api/*`) alongside the Vite-served browser app. Without
this, the browser-side `createRemoteTransport` had nothing to talk to during
local development.
Adds under `clients/web/server/`:
- vite-hono-plugin.ts (Hono middleware on the Vite dev server; gated to
`apply: 'serve'` and skipped when `process.env.VITEST` is set so it stays
inert in vitest/storybook runs)
- sandbox-controller.ts (standalone HTTP server for the MCP Apps sandbox)
- web-server-config.ts (env parsing + initial-config payload + banner;
`autoOpen` defaults off under VITEST so test runs don't shell out to `open`)
- server.ts (standalone Hono prod server for the future v2 launcher, #1246)
- start-vite-dev-server.ts (in-process Vite starter for the future launcher)
- vite-base-config.ts (shared optimizeDeps exclusions)
Wires `honoMiddlewarePlugin(buildWebServerConfigFromEnv())` into
`vite.config.ts` with `apply: 'serve'`. Adds the `open` dependency and a
DOM lib + `@inspector/core/*` path alias to `tsconfig.node.json` so the
dev-backend files type-check alongside vite.config.ts.
Coverage: web-server-config.ts (100% lines), sandbox-controller.ts (91% lines),
and vite-base-config.ts (100%) all clear the per-file gate. The Vite plugin,
prod server, and dev-server starter are excluded from coverage as runtime
glue exercised by `npm run dev` end-to-end (callers covered separately;
their logic lives in core/mcp/remote/node, which integration tests already
exercise).
Closes#1320.
Unblocks #1244 (Wire App.tsx) — its test plan can now run end-to-end.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(web/server): address PR review feedback on the dev-backend port
- sandbox-controller: resolve start() with empty values on listen error
(e.g. EADDRINUSE on a fixed MCP_SANDBOX_PORT) instead of leaving the
promise pending forever. Without this, the Vite plugin's
`await sandboxController.start()` could hang the entire dev backend on a
port collision. Banner now silently omits the sandbox line; getUrl()
keeps returning null. Adds a regression test exercising the EADDRINUSE
path against a real socket.
- vite-hono-plugin: listen for `error` and `close` on the incoming request
stream (in addition to `end`) so an aborted POST/PUT (browser navigates
mid-upload) doesn't leak the middleware indefinitely.
- vite-hono-plugin: add a `.catch()` to the initial `pump()` kickoff so a
sync throw before the first `await reader.read()` can't surface as an
unhandled rejection.
- sandbox_proxy.html: drop the stale CSP comment — v1.5's sandbox HTTP
server never set CSP either; this PR doesn't regress anything. Replace
with the actual isolation model (iframe sandbox attribute + origin
allowlist) so future readers don't assume a CSP header is enforced.
- Hoist `WebServerHandle` to a shared `server/types.ts` so the future
launcher port (#1246) doesn't need to pick one of two duplicate
declarations.
- Extract `resolveAutoOpen()` helper out of the triple-nested ternary in
`buildWebServerConfigFromEnv` for legibility, with the three-way logic
documented in one place.
- Add an inline comment explaining why
`webServerConfigToInitialPayload` falls back to streamable-http for
unknown discriminators (forward-compat with future SDK transports)
rather than throwing.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Hoist startRemoteServer(0) out of per-it() bodies into describe-scope
beforeAll/afterAll for the authentication, dangerouslyOmitAuth,
endpoint-error-paths, and Origin-validation describes — every it() in
these blocks only fires a single request and asserts response shape,
so one shared server per describe is safe.
Storage and additional-coverage describes are unchanged: their tests
mutate per-test filesystem state at common paths or use
heterogeneous server options, so sharing would require reorganizing
the suite (out of scope per the issue).
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* test: switch integration project to folder-based manifest (#1314)
Replace the `integrationTests` enumeration in vite.config.ts with a
folder glob (`src/test/integration/**/*.test.ts`). Files under that
folder run in the integration vitest project (node env, 30s timeouts);
everything else runs in the unit project (happy-dom).
This removes the silent misclassification trap from #1307: a new
integration-style test no longer falls into the unit project (and
happy-dom) just because someone forgot to add it to the array.
The 13 ported v1.5 tests + their helpers move under src/test/integration/
in a strict mirror of core/ source layout (mcp/, auth/, auth/node/,
storage/). Renames use git mv so history is preserved.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* test: address PR review on #1318 (glob extensions + AGENTS doc)
- vite.config.ts: widen integrationGlob from `*.test.ts` to
`*.test.{ts,tsx}` so the integration folder fully owns its contents.
Without this, a `.test.tsx` placed under src/test/integration/ would
match the unit project's `{ts,tsx}` include, fail the unit `.ts`-only
exclude, and silently land under happy-dom — the exact misclassification
trap this PR set out to close.
- AGENTS.md: update the line describing where integration tests live;
it still pointed at the old `src/test/core/` location and the removed
`integrationTests` enumeration.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* test(core): backfill coverage on inspectorClient + remote server
Resolves#1310. Drops the two TODO(#1310) exclusions from
clients/web/vite.config.ts (coverage.exclude); both files now pass the
per-file gate (lines >=90, statements >=85, functions >=80, branches >=50).
core/mcp/remote/node/server.ts (88.15 / 78.37 -> 97.29 / 100):
- Extract createTokenAuthProvider to its own file
(core/mcp/remote/node/tokenAuthProvider.ts) so the 7 no-op
OAuthClientProvider stubs can be unit-tested directly.
- Add focused integration tests covering the remaining gaps:
* createTransportNode failure path on POST /api/mcp/connect
* /api/mcp/send when transport is dead
* /api/log non-JSON body (catch fallback)
* forwardLogEvent unknown-level / no-messages / string-first-message
fallbacks (lines 232, 240-242, 257-265 in the original)
* /api/fetch missing-url + upstream-throw paths
* /api/storage POST invalid storeId + invalid JSON body
* /api/storage GET when the on-disk store is unparseable
core/mcp/inspectorClient.ts (73.48 / 72.09 -> 91.16 / 89.92):
Added a single "Coverage backfill (#1310)" describe block to the existing
integration test, plus a few unit-style cases:
- OAuth getters / setOAuthConfig / authenticate when oauthManager is unset.
- Simple session, roots, subscription, and instructions accessors.
- subscribe / unsubscribe error paths against a server without the
subscribe capability; happy path against one with it.
- getPrompt + readResourceFromTemplate happy path + expansion-failure.
- setLoggingLevel "Server does not support logging" guard.
- getAppRendererClient (null before connect, cached proxy after).
- OAuth + stdio: authenticate() throws via getServerUrl()'s stdio check.
- listChanged notifications: tools/resources/prompts handlers via the
fixture add_tool / add_resource / add_prompt tools.
- Misc tiny branches: re-connect no-op, elicit form mode, getTaskCapabilities
on a tasks-less server, receiver-task TTL cleanup + cancel via the
private surface (the focused-unit pass the issue describes).
- defensive `if (!this.client) throw` guards: one nullify-and-call test
exercises 16 of them via a single instance, plus the connect() entry
guard.
The receiver-task internals test uses (instance as unknown as
{ private fields... }) to drive createReceiverTask / cancelReceiverTask
/ TTL cleanup directly; per the issue, refactoring private helpers to
be testable from a unit test was authorized.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* test(core): address #1317 PR review feedback
- Rename the misnamed "listChanged notifications" describe in
inspectorClient.test.ts to "capability detection after connect", since
the test asserts only on getCapabilities() round-tripping and never
triggers or observes a notification dispatch. Real dispatch coverage
lives in the adjacent "list_changed + resourceUpdated notifications"
block. Kept the resources/prompts fixtures with a note explaining why
they are required for the server to advertise the matching capabilities.
- Expand the comment on the 250ms wait in remote-transport.test.ts'
/api/mcp/send dead-transport test so it is clear the >= 500 assertion
is intentionally permissive across both the dead-transport short-
circuit and the downstream send-failure branches.
- Add a "keep this in sync" pointer on the nullify() helper so a future
cleanup that removes any of the defensive `if (!this.client) throw`
guards in inspectorClient.ts is reminded to drop the matching call here
rather than silently under-covering.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* refactor(core): apply v1.5-port cleanups deferred from #1306
Addresses issue #1309 — five small, mechanical cleanups deferred from the
v1.5 runtime port:
1. Collapse BrowserOAuthStorage / NodeOAuthStorage / RemoteOAuthStorage
onto a shared OAuthStorageBase. Three ~130-line duplicates become
thin adapter wrappers; behavior + public API unchanged.
2. Re-enable erasableSyntaxOnly in tsconfig.app.json by rewriting the
seven (plus two the issue missed in sampling/elicitation create-
message) parameter-property constructors as explicit field + assign.
3. Route the lone console.error in inspectorClient.ts through this.logger.
4. Replace `${Date.now()}-${Math.random()}` with crypto.randomUUID() at
the five local-only ID sites (message tracking + sampling/elicit).
5. Pin the remote-server hono app generic at new Hono<Env>() and drop
the streamSSE cast. Root cause of the cast was an unresolved hono
subpath: the tsconfig "hono/*" wildcard didn't honor the exports
map, so hono/streaming leaked to the root-hoisted hono@4.12.12 while
hono itself resolved to clients/web/hono@4.12.18. Replaced the
wildcard with an explicit hono/streaming entry in both tsconfig.app
and tsconfig.test, removing the package-identity mismatch.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* refactor(core): address #1316 PR review feedback
- Drop the stranded no-op getServerState() read at the top of
clearClientInformation in OAuthStorageBase (carried over verbatim
from the three originals).
- Drop the crypto.randomUUID() guard at the task-ID site so all
six in-scope sites are consistent; the fallback branch was dead
code on every supported runtime.
- Remove the residual `c as Context` cast at the streamSSE call site
in core/mcp/remote/node/server.ts — once both hono imports
resolve to the same package the route-typed Context flows through
to streamSSE with no cast at all.
- Move the new `options` / `config` field declarations in
RemoteClientTransport up with the rest of the field block, matching
the layout used by the other five rewritten classes in this PR.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat(test): run v1.5-ported integration tests under node-env vitest (#1307)
Adds an `integration` vitest project alongside `unit` and `storybook` so
the 13 v1.5-ported tests that need real HTTP/stdio servers (or fail under
happy-dom's SDK mock resolution) run again under node env with 30s
timeouts. Restores test-servers/ as a buildable package so the stdio
subprocess spawns a real .js file. Drops the v1.5-source `coverage.exclude`
globs and fills coverage gaps across the v1.5-ported core/ modules; two
large files (inspectorClient.ts, remote/node/server.ts) keep narrow
per-file excludes tracked in #1310.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* test(core): address PR #1312 review — TS6-safe tsconfig, real transport_error path
- Drop `baseUrl` / `paths` / `typeRoots` / `declaration*` from
test-servers/tsconfig.json. They were only needed for type
resolution, and we already pass `--noCheck` from `test-servers:build`.
Under TS 6+ the deprecated `baseUrl` exited with code 2, which
silently skipped vitest in the `&&` chains of test:integration /
test:coverage / validate.
- Rework the "send() throws after closed" test to drive `closed` via
an actual transport_error SSE event instead of mutating private
state through a cast — matches the real path the branch guards.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* wip(core): bulk-copy v1.5 InspectorClient runtime files (#1302)
Verbatim ports from origin/v1.5/main into v2/main core/:
- logging/ (2 files)
- storage/ + storage/adapters/ (4 files)
- auth/ + auth/browser/ + auth/node/ + auth/remote/ (16 files)
- mcp/config.ts + mcp/node/ (4 files, stdio transport)
- mcp/remote/ + mcp/remote/node/ (12 files, HTTP/SSE transport)
- mcp/inspectorClient.ts (~2169 LOC), messageTrackingTransport,
fetchTracking, oauthManager, index, state/index
Total: ~7,800 LOC of source files, copied unmodified via
`git show origin/v1.5/main:<path> > <dest>`.
Also:
- Add deps: pino, hono, @hono/node-server, atomically, zustand, @types/node
- tsconfig.app.json: extend paths to resolve new node-only deps from
core/, add "node" to types, disable erasableSyntaxOnly (v1.5 code
uses parameter properties — TODO to re-enable in follow-up)
Compiles down to 47 type errors, all in three categories:
- 24 TS2305: v2 dropped v1.5 type exports (InspectorClientOptions,
CreateTransport, InspectorClientEnvironment, AppRendererClient,
SamplingCreateMessage class, ElicitationCreateMessage class,
jsonUtils helpers, OAuth events). Per user direction, will re-add
these to v2's diverged files (types.ts, samplingCreateMessage.ts,
etc.) so the ported code stays verbatim.
- 15 TS7006: implicit-any params in v1.5 code (v1.5 tsconfig was
looser than v2's noImplicitAny). Need explicit annotations.
- 8 other TS errors: ../package.json import, Transport|null narrowing,
hono Context generic mismatch, one unused parameter.
Tests not yet ported (19 files, ~11,170 LOC).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* wip(core): integrate ported v1.5 InspectorClient + port tests (#1302)
Source integration (typecheck + build now pass):
- Re-add v1.5 types/classes dropped during v2 evolution:
- types.ts: CreateTransport*, AppRendererClient, InspectorClient{Environment,Options}
- samplingCreateMessage.ts / elicitationCreateMessage.ts: classes + keep v2's
InspectorPending{Sampling,Elicitation} interface
- inspectorClientEventTarget.ts: oauth* event keys
- json/jsonUtils.ts: convertParameterValue / convertToolParameters /
convertPromptArguments + ParameterSchema
- inspectorClient.ts: replace `../package.json` import (v2 has no
core/package.json) with a hardcoded identity constant
- remote/node/server.ts: cast hono Context for streamSSE (hono's
generic typing tightened since v1.5)
- oauth-callback-server.ts: drop unused hostname field
- tsconfig: enable @types/node, add paths for new bare-module deps
(pino, zustand, hono, atomically) in app + test configs
- vite.config.ts: mirror those module aliases for the unit project
(which runs from repoRoot)
Tests ported (19 files, ~11K LOC) under clients/web/src/test/core/:
- All core/__tests__/auth/* (7 files)
- inspectorClient.test.ts + 4 OAuth variants
- mcp/oauthManager.test.ts, jsonUtils.test.ts
- transport, remote-transport, remote-server-config, storage-adapters
- helpers/oauth-client-fixtures.ts
Bulk-rewrote `from "../mcp/…"` → `from "@inspector/core/mcp/…"` etc.
Current state:
- typecheck: 0 errors
- build: passes (vite v8, 788 modules)
- unit tests: 1014 / 1034 pass (937 pre-existing + 77 new core tests)
- 20 failures in 13 files; failure modes:
- 13 file-load failures needing the v1.5 `test-servers/` package
(@modelcontextprotocol/inspector-test-server) which isn't in v2 yet
or real fs/network not available in happy-dom
- 7 assertion failures in discovery.test.ts + state-machine.test.ts
(SDK 1.29 vs 1.25 API drift — mocks don't match new shape)
Coverage gate not yet checked against new files.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* wip(core): port test-servers, more types, partial test integration (#1302)
- Port v1.5 test-servers/ package (12 files, ~5K LOC) under
repo-root/test-servers/, wire @modelcontextprotocol/inspector-test-server
alias in both vite.config.ts (regex form) and tsconfig.test.json.
- Add deps: express ^5, yaml ^2, @types/express ^5.
- Re-export InspectorClientOptions/Environment/Transport* from inspectorClient.ts
so v1.5 tests that import them from that module path keep resolving.
- Switch InspectorClientEventMap to use SamplingCreateMessage /
ElicitationCreateMessage classes (matches v1.5 event payload contracts)
while keeping the InspectorPending{Sampling,Elicitation} interfaces in
the *.ts modules for v2 consumers that still use them.
- Fix verbatim-module-syntax issue in test-servers and drop unused stdio
config field.
- Adjust inspectorClient.test.ts task types to TaskWithOptionalCreatedAt
(SDK 1.29 tightened Task shape vs v1.5's 1.25).
- Drop spurious maxMessages/maxFetchRequests/maxStderrLogEvents from
remote-transport.test.ts (never on InspectorClientOptions).
State:
- typecheck: 0 errors
- build: passes
- unit tests: 1099 / 1270 pass
- 171 failures in 10 test files, almost all integration tests that need
real HTTP/fs/network (happy-dom can't host them). Next: add a node-env
vitest project to run these properly.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat(core): port v1.5 InspectorClient + transports + auth subsystem (#1302)
Bulk-port of the v1.5 core runtime (~7,800 LOC source + ~16K LOC tests +
~5K LOC test-servers/) into v2. Files copied verbatim from origin/v1.5/main
via `git show >`; only minimal adapters added on the v2 side.
Subsystems landed
-----------------
- core/logging/, core/storage/, core/auth/ (browser + node + remote)
- core/mcp/ runtime: inspectorClient.ts (2169 LOC), oauthManager,
messageTrackingTransport, fetchTracking, config, index
- core/mcp/node/ (stdio transport)
- core/mcp/remote/ + remote/node/ (HTTP/SSE transport + Hono server)
- test-servers/ package (composable MCP test servers, fixtures, harness)
v2 adapter layer
----------------
- types.ts: append CreateTransport*, AppRendererClient,
InspectorClient{Environment,Options} (v2 had pruned these — v1.5 needs them)
- samplingCreateMessage / elicitationCreateMessage: keep v1.5 classes AND
v2's Inspector Pending* interfaces in the same files
- inspectorClientEventTarget: add oauth* event keys; switch pending
sampling/elicitation payloads to the classes (matches v1.5 contracts)
- json/jsonUtils: append convertParameterValue/convertToolParameters/
convertPromptArguments + ParameterSchema
- inspectorClient.ts: re-export Options/Environment/Transport* so tests
that import these from `inspectorClient.js` keep resolving; replace
`../package.json` import with a hardcoded identity (v2 has no core/package.json)
- remote/node/server.ts: cast hono Context for streamSSE (hono generic
tightened since v1.5)
- oauth-callback-server: drop unused hostname field
- tsconfig.app.json / tsconfig.test.json: add @types/node + paths for
pino/zustand/hono/@hono/node-server/atomically/express/yaml, disable
erasableSyntaxOnly (parameter-property syntax from v1.5)
- vite.config.ts: regex-form module aliases (so the package's exports
field handles subpaths like `hono/streaming`, `zustand/middleware`),
alias for @modelcontextprotocol/inspector-test-server → test-servers/src
Tests
-----
- 19 v1.5 test files copied under clients/web/src/test/core/,
imports rewired to @inspector/core/* and @modelcontextprotocol/inspector-test-server
- 1004 unit tests pass; 298 storybook tests pass; coverage gate passes
Known follow-ups (out-of-scope here, tracked in TODOs)
------------------------------------------------------
- 13 integration test files are excluded from the unit project (spawn real
HTTP/stdio servers; run real OAuth flows; mock SDK auth → happy-dom CORS).
Needs a node-env vitest project. Files listed under `test.projects[0].test.exclude`.
- Source files exercised only by those integration tests are excluded from
the coverage gate pending the node-env project. Listed under
`test.coverage.exclude` with comment.
- Manual smoke tests against real stdio / HTTP / OAuth server per the AC
on #1302 still need verification locally.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* docs/fix: address #1306 review (AGENTS tree, #1307 links, @types/node, .substr)
- AGENTS.md: expand the project-structure tree to include all v1.5-ported
subsystems (core/auth/{browser,node,remote}, core/storage, core/logging,
core/mcp/{node,remote/{,node}}, core/mcp/state) plus the new top-level
test-servers/ directory (with a note about its @modelcontextprotocol/
inspector-test-server alias).
- vite.config.ts: reference #1307 from both the test exclude block and the
coverage exclude block so the next person editing either sees the
marching order for shrinking the lists.
- package.json: restore @types/node to ^24 (an earlier `npm install
--save-dev @types/node@^22` accidentally pinned it lower than the repo
was using; nothing here required ^22).
- core/mcp/fetchTracking.ts: replace deprecated .substr(2, 9) with
.slice(2, 11) to match the convention in core/mcp/inspectorClient.ts.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(coverage): include core/ files in v8 per-file gate (#1301)
Two issues prevented coverage.include's '../../core/...' globs from
reaching core/ files:
1. allowExternal defaults to false — v8 drops any file whose absolute
path doesn't start with the project root (clients/web/) before the
include filter runs. Set allowExternal: true, and switch the globs
to absolute paths so tinyglobby + picomatch handle them correctly.
2. vitest's coverage transformer only processes files inside a project
root; for everything else it falls back to raw-TS parsing in
rolldown, which can't handle TS-only syntax (e.g. import type) —
silently dropping untested core/ files and bypassing the gate.
Root the unit project at the repo root so the transformer reaches
core/. The repo root has no node_modules, so add an explicit react
alias and shift include/setupFiles paths accordingly.
Also add the four pure-type core files (samplingCreateMessage,
sessionStorage, inspectorClientProtocol, inspectorClientEventTarget)
to the exclude list with a comment — they have no executable code.
Verified by introducing an untested probe file, confirming exit=1 with
threshold errors, then removing it.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* refactor(coverage): address review feedback on #1305
- Drop allowExternal: true — with the unit project rooted at repoRoot,
core/ files are no longer "external" to any project that collects
coverage (the storybook project doesn't), so the flag is redundant.
- Extract sharedAliases / sharedDedupe constants and reference them
from both the top-level resolve and the unit project resolve, so
the two stay in sync if a new core/* alias is added.
- Add TODO(#1243) on the inspectorClientEventTarget.ts exclusion so
it's revisited once the class gains real behavior in the
InspectorClient port.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat(core): port v1.5 hook layer (useInspectorClient + useManaged*/usePaged* + log hooks)
Wholesale port of v1.5's React hook layer and the EventTarget-based state machinery the hooks subscribe to (#1243). Hooks decouple from the concrete InspectorClient via a new `InspectorClientProtocol` interface so this layer can land before the runtime class is ported.
Added under `core/`:
- `core/react/` — useInspectorClient + 5 useManaged* + 5 usePaged* + 3 log hooks
- `core/mcp/state/` — 5 managed*State + 5 paged*State + messageLogState + stderrLogState + fetchRequestLogState
- `core/mcp/inspectorClientProtocol.ts`, `inspectorClientEventTarget.ts`, `typedEventTarget.ts`, `samplingCreateMessage.ts`, `sessionStorage.ts`
- v2-only wrapper types in `core/mcp/types.ts`: InspectorUrlElicitRequest, InspectorPendingRequest, InspectorRequestHistoryItem
Tests live under `clients/web/src/test/core/` (28 new test files, 264 new tests). Test fake: `core/mcp/__tests__/fakeInspectorClient.ts`.
Build wiring:
- `vite.config.ts`: dedupe react/react-dom so core/ files resolve them via clients/web/node_modules.
- `tsconfig.app.json` / `tsconfig.test.json`: paths map react/react-dom/vitest to clients/web/node_modules so `tsc -b` resolves bare imports from core/ source files (which sit above the package's node_modules). Exclude core/__tests__ from app config; include in test config.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* review: address PR #1300 review findings
Bugs:
- fetchRequestLogState: add .catch() on sessionStorage.loadSession so a corrupt
or unreadable session doesn't surface as an unhandled rejection (matches the
saveSession swallow pattern).
- messageLogState: sync pendingRequestEntries when clearMessages() evicts a
pending request entry, so a later response with the same id falls through
to the append branch instead of mutating an unreachable map entry. Add tests
for both predicate and no-predicate paths.
Cleanups:
- Extract mergeTaskIntoList helper to core/mcp/state/mergeTaskIntoList.ts so
managed/paged RequestorTasks state classes share one definition.
- Add a one-line JSDoc note on each PagedTools/Prompts/Resources/Templates
state explaining why they deliberately do not subscribe to *ListChanged
(cursor invalidation mid-pagination).
- Document on useInspectorClient that appRendererClient is read lazily on
every render and is not subscribed (changes once at connect time).
Carry-overs from v1.5 deliberately left for the follow-up InspectorClient
runtime port:
- Refresh/status-change race in Managed*State.refresh() (status check is only
at entry; a disconnect during await can re-clear before refresh resumes).
- nextCursor exposure asymmetry across usePaged* (only usePagedRequestorTasks
surfaces it via state field; others return it from loadPage() only).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* ci(storybook): wire test:storybook into CI (#1251)
- Add test:storybook:watch to clients/web for interactive iteration; the
one-shot script already existed.
- Surface test:storybook at the repo root, delegating to the web client.
- Add a separate CI step that installs the Playwright Chromium binary
and runs the storybook play-function suite after validate. Kept out of
validate because the suite needs a real browser and is ~10s slower
than the unit tests.
- Document the new pre-push step in AGENTS.md.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* ci(storybook): cache Playwright browsers + tighten AGENTS wording
Cache the Chromium binary keyed on clients/web/package-lock.json so the
~150MB download is a no-op on CI cache hits. Also clarify that CI runs
test:storybook after the unit/lint/build checks (not literal `validate`,
which is a local-only script).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat(apps): render tool.icons on ToolListItem (#1264)
Add icon rendering to ToolListItem so the Tools sidebar matches the
Apps sidebar treatment. Falls back gracefully when tool.icons is absent.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(stories): use #228be6 in ToolListItem WithIcon data URL
The source SVG had %23228be6 pre-encoded, but encodeURIComponent then
re-escaped the % to %25, leaving fill="%23228be6" in the decoded SVG
and making the icon render black instead of Mantine blue.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat(apps): render tool.icons on ToolDetailPanel (#1264)
Show the tool icon next to the title in the detail panel, matching the
header icon size used in AppsScreen (24x24, fit: contain). Adds a
WithIcon story and tests for icon-present/icon-absent states.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat(apps): AppsScreen and InspectorView wiring (#1263)
Adds the Apps tab as a top-level screen composed of `AppListItem`,
`AppDetailPanel`, and `AppRenderer`. The screen is presentational —
selection, form values, running, and maximize state are local; data and
side effects come in via props and callbacks.
InspectorView filters `tools` through `isAppTool` (wrapped in a try/catch
so a malformed `_meta.ui.resourceUri` skips one tool instead of dropping
every following App), exposes the Apps tab in `ALL_TABS` after Tools, and
wires Phase-3 stub callbacks plus a stub bridge factory.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* refactor(apps): extract AppControls and rename Refresh button
Pulls the sidebar list, search input, "List updated" indicator, and
Refresh button out of `AppsScreen` into a dedicated `AppControls`
group, mirroring `ToolControls` and `TaskControls`. The Refresh button
moves from a full-width row beneath the list to the top-right of the
title row (single label "Refresh", matching `TaskControls`).
`AppsScreen` keeps the right-pane orchestration (selection, form,
running, maximize, header actions) and now delegates the entire
sidebar to `AppControls`.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* chore(stories): seed InspectorView demo tools with MCP Apps
Adds three apps (Cohort Data, Weather Widget, Ops Dashboard) carrying
real `_meta.ui.resourceUri` to the demo tools list so the Apps tab has
content in the Default story alongside the existing tools.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(apps): drop AppDetailPanel title row to avoid duplicate name
The parent AppsScreen header already shows the app name, so the panel's
own title/icon row rendered the name twice. Remove the title row (and
its now-unused imports, styled subcomponents, and the four tests that
asserted on title/icon rendering).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* refactor(resources): drop annotation badges from ResourceListItem
The list row now renders just the resource title/name. Audience and
priority badges remain visible in the preview/template panels, so the
data is still surfaced — just not duplicated in the picker. Drops the
now-unused AnnotationBadge import, the badge-only stories, and the
badge-rendering test.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* chore(apps): address review — share hasInputFields, show header icon
- Extracts the duplicated `Object.keys(inputSchema.properties ?? {}).length > 0`
check from `AppsScreen` (and the equivalent in `AppDetailPanel`'s
`hasFields`) into a shared `hasInputFields` helper in `toolUtils`,
with tests.
- Renders the selected tool's first icon next to the right-pane header
title in `AppsScreen`, restoring the icon visibility that was lost
when `AppDetailPanel`'s title row was removed.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* chore(apps): finish hasInputFields extraction; document isOpening
- AppDetailPanel now actually calls the shared `hasInputFields` helper
instead of inlining the same `Object.keys(...).length > 0` check.
- AppsScreen comments why `isOpening={false}` is hardcoded: the screen
flips to AppRenderer synchronously, so the loading state is unreachable
here — the prop remains for standalone use and Phase 3 wiring.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat(apps): add AppRenderer element with imperative ref (#1262)
Introduces a thin host-side shell for embedded MCP Apps. The renderer
owns the iframe element and the AppBridge lifecycle; the wiring layer
constructs the bridge from the active MCP Client via bridgeFactory and
drives input/result/cancellation through the imperative ref handle.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* review: address PR feedback on AppRenderer
- Document BridgeFactory stability requirement (unstable identity thrashes
the iframe on every render).
- Null bridgeRef synchronously inside teardown() before awaiting disposal,
so a concurrent unmount cleanup cannot dispose the same bridge twice.
- Comment the iframe trust model: the outer frame loads the inspector's
trusted same-origin sandbox-proxy, which sandboxes the actual app
content. Adding `sandbox` to this frame would block the postMessage
bridge.
- Rename `Error` story export to `BridgeError` to avoid shadowing the
global Error constructor; drop the `globalThis.Error` workaround.
- Comment flushAsync to explain the two-tick microtask budget.
- Add unmount-while-tearing-down test to lock in the no-double-dispose
guarantee.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat(apps): add AppListItem and AppDetailPanel groups (#1261)
Two presentational groups for the upcoming AppsScreen.
AppListItem (clients/web/src/components/groups/AppListItem/):
- Props { tool, selected, onClick }
- Renders tool.icons[0] (sized sm) when present, label (title ?? name),
two-line clamped description, and a trailing chevron affordance.
- Reuses ToolListItem's selected-state highlight (UnstyledButton bg via
--mantine-primary-color-light).
AppDetailPanel (clients/web/src/components/groups/AppDetailPanel/):
- Props { tool, formValues, isOpening, onFormChange, onOpenApp }
- Layout: optional icon + title row, optional description, divider,
SchemaForm bound to tool.inputSchema, full-width "Open App" button.
- "Open App" is disabled while isOpening or while any required input
field is empty (validation derived from tool.inputSchema.required).
- No annotations, progress, or cancel — execution is delegated to the
embedded app, not run as an MCP tool call from the panel.
Convention notes:
- Issue text references @tabler IconChevronRight / IconPlayerPlay; the
v2 web client already uses react-icons (md/ri namespaces), so this PR
uses MdChevronRight and MdPlayArrow for consistency rather than
introducing a second icon library.
Stories: AppListItem (Default, Selected, WithIcon, NoDescription, LongName);
AppDetailPanel (NoFields, SimpleStringParam, MultipleParams, WithIcon,
Opening, ComplexSchema).
Tests: 22 cases covering rendering branches, callback wiring, and the
disabled-button derivation. Per-file coverage 100% across lines,
statements, functions, and branches for both new components.
Closes#1261
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(stories): inline SVG for WithIcon stories so they render reliably
The Wikipedia-hosted PNG used in the AppListItem and AppDetailPanel
WithIcon stories returned a broken image in Storybook (rate-limited /
hot-link blocked). Replace it with an inline SVG sun icon as a data
URI so the stories render offline, in Chromatic, and on any network.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* test(apps): address PR #1273 review feedback
- Extract resolveDisplayLabel(name, title?) to src/utils/toolUtils.ts and
migrate all four call sites (AppListItem, AppDetailPanel, ToolListItem,
ToolDetailPanel) to consume it. Avoids the fourth copy when the next
list/detail screen arrives. Includes a small unit suite.
- AppDetailPanel: only render the <Divider/> when the schema has at least
one property — previously a no-input app showed an orphaned separator
between the description and the Open App button.
- AppDetailPanel: align PanelTitle truncate to `true` (matches the
ItemLabel form used in AppListItem / ToolListItem); behaviorally
identical, just less surprising for future readers.
- AppDetailPanel: add an inline comment explaining that SchemaForm is
disabled only while isOpening, not while validation fails — users
should be able to keep filling required fields. The validation gate
lives on the Open App button.
- Extract the inline SUN_ICON_SVG data URI used by the WithIcon stories
to src/test/fixtures/storyIcons.ts so the next story that needs it
doesn't grow a third copy.
- Tests: add the AppDetailPanel divider-conditional cases (rendered
with fields, omitted without). Per-file coverage stays at 100% for
AppListItem, AppDetailPanel, ToolListItem, ToolDetailPanel, and the
new toolUtils. Suite: 70 files / 618 tests passing.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* test(toolUtils): rename empty-string test for readability
Drops the doubled "is ... is preserved" phrase in the test name.
Behavior unchanged.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat(apps): add isAppTool helper in core (#1260)
Adds @modelcontextprotocol/ext-apps as a dependency (root + clients/web)
and a thin core/mcp/apps.ts wrapper exposing:
- getAppResourceUri(tool): re-export of getToolUiResourceUri so all
Inspector clients consume the same nested/flat _meta resolution.
- isAppTool(tool): boolean predicate built on top of getAppResourceUri.
Both surface the underlying ext-apps throw on a malformed _meta.ui.resourceUri
(string that does not start with "ui://") rather than swallowing it — server
bugs should be visible.
Tests cover nested format, deprecated flat format, nested-wins precedence,
missing _meta, and the invalid-URI throw.
Refs #1260
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* test(apps): address PR #1272 review feedback
- core/mcp/apps.ts: convert getAppResourceUri from a typed const re-export to
a function declaration so its export style matches isAppTool. Behavior is
unchanged.
- core/mcp/apps.ts: document the filter-iteration implication on isAppTool —
callers using tools.filter(isAppTool) will halt on a server-malformed URI.
- apps.test.ts: add the symmetric isAppTool false-case for `_meta.ui` present
without `resourceUri`.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* test(coverage): bootstrap vitest unit infra and add tests at 90%+ per file (#1245)
Establishes the unit-test runtime AGENTS.md mandates: a vitest "unit"
project alongside the existing storybook project, RTL + jest-dom +
happy-dom, a Mantine-aware render helper, and a 90%-line per-file
coverage gate enforced by CI. Adds 66 test files covering all 62
components under elements/groups/screens/views, plus core/mcp and
clients/web/src/utils. Total: 577 tests, 98.7% line coverage globally.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* test(coverage): address PR #1269 review feedback
- AGENTS.md: correct coverage gate description (lines ≥ 90, statements ≥ 85,
functions ≥ 80, branches ≥ 50) with rationale for the relaxed branch
threshold
- InspectorView.test.tsx: document what the Math.random() mock values
(0.1 / 0.99) control vs. STUB_SUCCESS_RATE in InspectorView.tsx
- vite.config.ts / tsconfig.test.json: drop unused `globals: true` and the
`vitest/globals` type now that every test file imports describe/it/expect/vi
explicitly; manual `afterEach(cleanup)` in setup.ts remains required because
testing-library auto-cleanup relies on globalThis.afterEach
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The ResourceSelected story passed both sampleResources and
sampleSubscriptions, each of which renders 'config.json' as text. The
play helper's canvas.findByText('config.json') matched two nodes and
rejected.
Extends clickByText to accept an optional regionName regex; ResourceSelected
now scopes the click to within the 'URIs' Accordion panel by aria-region
name, so the subscriptions panel's matching label is filtered out without
changing what the story demonstrates.
Full storybook suite now green (273/273), unblocking #1251.
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Cliff Hall