Closes#1577
Extends the deep link (#1576) to land on a rendered MCP App:
- `&openApp=<toolName>` — once connected and the tool appears in the app list,
InspectorView switches to the Apps tab and pre-selects it.
- `&appArgs=<base64url(JSON)>` — pre-fills the tool form; values are merged
**over** `collectSchemaDefaults()` so required-with-default fields don't
disable "Open App".
- `&autoOpen=<token>` — token-gated auto-click of "Open App".
Adapted to the current architecture (differs from PR #1510's reference):
- Tab state is lifted to App.tsx (#1417), so the openApp effect lives in
InspectorView and drives `onActiveTabChange("Apps")` + `onAppsUiChange` seed
+ `onSelectApp`, guarded by `deepLinkOpenAppRef` to fire once.
- AppsScreen owns the `running`/iframe state, so `autoOpen` is threaded to it
as a prop; a fire-once effect defers the open one microtask past the
synchronous effect body (satisfying `react-hooks/set-state-in-effect`,
matching App.tsx's connect effect). `resetAppChannels` memoized so the
effect's deps don't churn.
Tests: InspectorView (tab-switch + pre-select + appArgs-over-defaults merge +
non-app openApp ignored), AppsScreen (autoOpen fires once with seeded values;
no-selection stays idle). README documents the params. Coverage gate green.
Part of the #1579 decomposition (Wave 4).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BrihGWcrM9JGRyu41nzZYw
* web: sandbox hardening — opaque-origin inner frame, srcdoc-only, frame-ancestors
Harden the MCP Apps sandbox isolation model:
- The inner (untrusted-widget) iframe is sandboxed WITHOUT `allow-same-origin`
(default "allow-scripts allow-forms") and any server-supplied sandbox value
has `allow-same-origin` stripped case-insensitively, so the widget runs under
an opaque origin and cannot reach the proxy's DOM to bypass its own CSP via
the parent realm.
- Widget delivery is `srcdoc`-only (no document.write), which respects the
opaque origin.
- The relay accepts the inner frame's "null" origin with
`event.source === inner.contentWindow` as the unforgeable identity check.
- sandbox-controller.ts serves the proxy with
`Content-Security-Policy: frame-ancestors http://127.0.0.1:* http://localhost:*`;
fetch directives are deliberately omitted so a srcdoc iframe's inherited
policy container does not override the per-app CSP.
Updated the isolation-model comments and the sandbox-controller integration
test asserts the frame-ancestors header (and the absence of fetch directives).
Closes#1565
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BrihGWcrM9JGRyu41nzZYw
* web: enforce per-app _meta.ui.csp in the app bridge + surface resource-read failures
- createAppBridgeFactory now builds each app's `_meta.ui.csp` host-side via the
sandbox-csp library (approveCspSources → buildSandboxCspPolicy) and wraps the
untrusted widget HTML with wrapSandboxedHtml so the CSP <meta> is always the
first <head> child before the HTML reaches the sandboxed iframe. The approved
(post-filter) csp + permissions are echoed back via hostCapabilities.sandbox
so the view sees what was granted. The csp is no longer sent inline through
sendSandboxResourceReady — it is enforced via the wrapped meta.
- A per-app copy of HOST_CAPABILITIES prevents the sandbox echo from mutating
the shared constant.
- New optional `onResourceError` dep on AppBridgeFactoryDeps routes resource
read/post failures to the caller (also always console.error'd) instead of
leaving a silent blank frame.
Closes#1566
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BrihGWcrM9JGRyu41nzZYw
* web: full hostContext delivery + live host-context-changed updates
- createAppBridgeFactory now seeds the full hostContext via
snapshotHostContext(iframe, HOST_AVAILABLE_DISPLAY_MODES): theme, styles
(Mantine design tokens), displayMode ("inline"), availableDisplayModes, and
containerDimensions — replacing the theme-only seed. The local currentTheme
helper is dropped in favor of hostContext.ts.
- AppRenderer gains `displayMode` and `containerRef` props and pushes live
host-context changes via the SDK's sendHostContextChange (partial
host-context-changed notifications), never a hand-maintained snapshot:
- theme + styles via a MutationObserver on <html data-mantine-color-scheme>,
- containerDimensions via a ResizeObserver on the host container (or the
iframe fallback), gated on `initialized`, skipping 0x0 and value-equal
repeats, plus a one-shot push on initialize once layout settles,
- displayMode whenever the prop changes (Maximize/Restore), gated on init.
Adds AppRenderer host-context tests (theme flip incl. resolved styles,
displayMode push + init gate, container-dimensions on init and resize, observer
teardown) and updates the factory + story mocks for the new hostContext shape
and sendHostContextChange method.
Closes#1567
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BrihGWcrM9JGRyu41nzZYw
* web: honor ui/notifications/size-changed and ui/request-display-mode
AppRenderer:
- New `onSizeChange` prop forwards the view's ui/notifications/size-changed
reports (via the bridge `sizechange` event).
- New `onRequestDisplayMode` prop backs an `onrequestdisplaymode` bridge handler
that returns the mode the host actually applied; with no handler the request
is declined by returning the current host-side mode (falling back to inline).
AppsScreen:
- Adds a host-controlled RendererContainer (measured for containerDimensions)
with an inner box sized by the view-reported height (`appHeight`), so
host→view container size and view→host size-changed don't couple.
- `handleSizeChange` records the reported height (ignored while maximized);
`displayMode` is derived from the maximize toggle and pushed to the view;
`handleRequestDisplayMode` filters against HOST_AVAILABLE_DISPLAY_MODES
(declining e.g. "pip") and maps fullscreen↔inline onto the maximize state.
Adds AppRenderer tests (size forwarding, request-display-mode apply/decline/
fallback) and AppsScreen tests (frame sizing on report, no-height ignore,
fullscreen/restore/decline via request-display-mode). Story mocks gain
sendHostContextChange/addEventListener.
Closes#1568
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BrihGWcrM9JGRyu41nzZYw
* web: honor ui/download-file with confirmation and http(s) allowlist
createAppBridgeFactory now advertises the `downloadFile` host capability and
attaches an `ondownloadfile` handler:
- Requires a host-mediated confirmation (window.confirm) listing the requested
file(s); a declined prompt or an empty payload returns isError without acting.
- Inline EmbeddedResources (text or base64 blob) are written to disk via the
shared downloadBlob object-URL anchor; base64 is decoded via base64ToBytes.
- ResourceLinks are opened in a new tab, restricted to http(s) by the shared
isHttpUrl allowlist (javascript:/data:/file:/malformed are rejected).
- Labels shown in the confirmation are sanitized (control + format chars
stripped, length clamped) so a server-supplied filename/URI can't inject
newlines or reflow the prompt.
- Partial-batch success is surfaced (isError only when nothing downloaded), with
skipped items warned.
Adds factory tests for the full download surface (text/blob/link, decline,
empty, throw, non-http rejection, label sanitization/clamping, partial batch).
Closes#1569
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BrihGWcrM9JGRyu41nzZYw
* web: surface ui/message and app logging in panels below the running app
AppRenderer:
- New `onMessage` prop backs an `onmessage` bridge handler that surfaces the
view's ui/message content and returns the spec-required empty result;
with no handler the submission is declined (isError).
- New `onLog` prop forwards MCP `notifications/message` log entries via the
bridge `loggingmessage` event, honoring the advertised `logging` capability.
AppsScreen:
- Adds `messages`/`appLogs` state (cleared with the reported height on
select/open/close/back via resetAppChannels) and `handleMessage`/`handleLog`.
- Renders a "Messages from app" panel (data-testid="apps-messages", reusing
ContentViewer) and a default-expanded collapsible "App logs" panel
(data-testid="apps-logs", reusing LogLevelBadge) with logger names and a
Clear button, both as pinned panels below the running widget.
Adds AppRenderer tests (log forward + no-handler no-throw, ui/message
route/empty-result/decline) and AppsScreen tests (message log render + empty
result + clear-on-close, log panel expand/collapse/clear).
Closes#1570
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BrihGWcrM9JGRyu41nzZYw
* web: stage partial input — replay tool-input-partial before final tool input
AppRenderer:
- New `partialInputs` prop (ordered fragments), snapshotted into pendingPartials
at bridge-build time (read via a ref so prop churn never rebuilds the iframe)
and cleared on dispose. flushPending replays them via
bridge.sendToolInputPartial BEFORE the complete tool-input, per spec.
AppsScreen:
- New `partialStages` state + `handleStagePartialInput` (snapshots current form
values). resetAppChannels gains a `keepPartials` option so Open App preserves
the staged fragments (they're consumed by the renderer) while select/close/
back clear them.
- Adds a "Stage partial input" control (fielded apps only) with a staged count
and "Clear staged", wired to AppRenderer's `partialInputs`.
Adds AppRenderer tests (ordered replay before tool-input; none when omitted) and
AppsScreen tests (stage/count/clear gated on fields; staged partials survive
Open and are replayed).
Closes#1571
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BrihGWcrM9JGRyu41nzZYw
* web: app lifecycle status + machine-readable data attributes on the Apps screen
AppRenderer:
- New AppRendererStatus ("loading" | "ready" | "error") + onAppStatusChange
callback. Fires "loading" at the start of every (re)build, "ready" on the
view's initialized signal, and "error" on a factory throw/rejection.
AppsScreen:
- appStatus/appError state; handleAppError captures the reason locally and
forwards to the parent onError. resetAppChannels clears both.
- Surfaces the contract as data-testid="apps-form" carrying data-app-status and
data-app-error, renders an error panel (data-testid="apps-error") in place of
the blank frame, and tags the stage button data-testid="apps-stage".
- AppDetailPanel's Open App button gains data-testid="open-app".
App.tsx:
- Wires the bridge factory's onResourceError to a red toast so a malformed/404
UI resource is no longer console-only (the renderer separately drives
data-app-status, so a driver times out on never-reaching-ready and reads it).
Documents the data-testid/data-* automation contract in clients/web/README.md.
Adds AppRenderer status-transition tests and AppsScreen status/error-panel
tests.
Closes#1572
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BrihGWcrM9JGRyu41nzZYw
* review: warn when server sandbox drops allow-scripts + revert incidental lockfile churn
Addresses the claude[bot] review on PR #1650:
- sandbox_proxy.html: log a console.warn when a server-supplied `sandbox`
value lacks `allow-scripts` after the `allow-same-origin` strip, so a
fail-closed (app-never-runs) sandbox is diagnosable instead of silent.
- Restore clients/{cli,launcher,tui,web}/package-lock.json to their
v2/main state — the earlier churn (@emnapi/@napi-rs optional deps, a
dropped version field) was incidental `npm install` regeneration
unrelated to the sandbox change.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BrihGWcrM9JGRyu41nzZYw
* review: document CSP-vs-permissions filtering asymmetry (PR #1652)
Address the claude[bot] review's finding #2: add a comment explaining why
`csp` is injection-filtered (approveCspSources) while `permissions` is passed
through verbatim — permissions is a structured boolean object consumed only by
the sandbox proxy's buildAllowAttribute() (fixed key→token mapping), so it
never reaches the iframe sandbox/allow attribute as raw text. No logic change.
Also merges v2/1565-sandbox-hardening (review fixes) up the stack.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BrihGWcrM9JGRyu41nzZYw
* test: lock in per-app HOST_CAPABILITIES isolation (PR #1652 round 2)
Add a test asserting the shared HOST_CAPABILITIES.sandbox stays undefined after
a sandboxready run sets hostCapabilities.sandbox — guards against a silent
regression if the per-app spread copy were ever dropped.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BrihGWcrM9JGRyu41nzZYw
* review: gate theme observer on initialized + document seed/dep assumptions (PR #1654)
Address the claude[bot] review:
- Finding #1 (applied): gate the theme+styles MutationObserver on the view's
`initialized` signal, matching the container/displayMode pushes — a theme
flip in the construction→handshake window no longer races ui/initialize. The
factory seeds the construction-time theme, and the first post-init flip
carries the current value. Added tests: pre-init flip is dropped; the live
flip test now completes the handshake first.
- Finding #2 (applied, comment): note in snapshotHostContext that the
`displayMode: "inline"` seed assumes inline-at-open (AppsScreen always opens
inline; #1568's displayMode push carries transitions).
- Finding #4 (applied, comment): explain why `containerRef` is a build-effect
dep yet doesn't force a rebuild (sameInputs ignores it; read lazily).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BrihGWcrM9JGRyu41nzZYw
* review: guard 0-height size reports, distinct RendererFrame, shared payload type (PR #1656)
Address the claude[bot] review:
- #1 (applied): handleSizeChange ignores a non-positive height — a view's
ResizeObserver can transiently fire 0 pre-layout/teardown, which would
collapse the frame; mirrors AppRenderer's 0×0 skip. Added a test.
- #2 (applied): reinstate a distinct `RendererFrame` constant for the inner
content-sized box so its role reads clearly vs. the outer RendererContainer.
- #3 (declined + comment): keep carrying `appHeight` across maximize→restore —
clearing it would flash the frame to full-card height until the next
size-changed; the carry-over restores at the prior size. Documented.
- #4 (applied): type handleSizeChange's param as AppBridgeEventMap["sizechange"]
so the payload shape has a single source of truth with AppRenderer's prop.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BrihGWcrM9JGRyu41nzZYw
* review: distinguish links, cap batch size, comment truncation (PR #1657)
Address the claude[bot] review:
- #1 (applied): the confirmation now reads "download or open N item(s)" and
resource_links are prefixed "↗" in the summary, so a link that opens in a tab
is distinguishable from an embedded file that saves to disk.
- #2 (applied): cap a single ui/download-file batch at MAX_DOWNLOAD_ITEMS (20);
an oversized batch is rejected (isError) before the prompt, with a warn.
- #3 (declined): can't check window.open's return value to detect a blocked
popup — we pass `noopener`, and per spec window.open returns null even on a
successful noopener open, so null can't distinguish blocked from succeeded.
- #4b (applied, comment): note the start-of-label truncation is intentional so a
link's scheme+host stays visible for the trust decision.
Adds tests: oversized batch rejected without confirming/acting; resource-link ↗
prefix in the summary.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BrihGWcrM9JGRyu41nzZYw
* review: skip embedded resource with neither text nor blob (PR #1657 round 2)
Address the round-2 observation: an untrusted EmbeddedResource carrying neither
`blob` nor a string `text` previously produced `new Blob([undefined])` — a file
containing the literal text "undefined", counted as success. Now such an item
is skipped (returns false, like a rejected link), so isError reflects it. Added
a test asserting no object URL is created and the batch reports isError.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BrihGWcrM9JGRyu41nzZYw
* review: type-honest formatLogData, cap channels, log-panel a11y (PR #1658)
Address the claude[bot] review:
- #1 (applied): formatLogData coalesces JSON.stringify(undefined) → "" so the
`: string` return type is honest for a data-less log.
- #2 (applied): cap retained messages/logs at MAX_APP_CHANNEL_ENTRIES (500) via
appendCapped — a chatty widget no longer grows the panels/DOM without bound
between Clear/close; oldest entries drop.
- #3 (declined): keep the Messages panel without a Clear/collapse — messages are
sparser and already cleared on close/switch; keeping it simple.
- #4 (applied): the App-logs toggle now has aria-controls pointing at the
Collapse region (id="apps-logs-region") for assistive tech.
Adds tests: no-data log renders (coalesce branch), 501 logs cap at 500 (oldest
dropped), aria-controls assertion.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BrihGWcrM9JGRyu41nzZYw
* review: stable id key for Messages panel (PR #1658 round 2)
Address the round-2 nit: give AppMessage a stable `id` (assigned via
nextMessageIdRef in handleMessage) and key the Messages panel on it instead of
the array index — so the appendCapped front-drop can't renumber keys once a run
exceeds the cap, matching AppLogEntry's treatment.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BrihGWcrM9JGRyu41nzZYw
* review: note partialStages intentionally survives handleOpen (PR #1659)
Address the claude[bot] review's observation #2: add a comment at the
handleOpen call site clarifying that partialStages is intentionally NOT cleared
there — AppRenderer snapshots the fragments into its own pendingPartialsRef at
build time, and the staging UI only renders while not running, so the surviving
state is invisible until the next reset drains it.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BrihGWcrM9JGRyu41nzZYw
* review: align error-panel docs with behavior + gate data-app-error on running (PR #1660)
Address the claude[bot] review:
- #1 (applied, docs): the error panel renders BELOW the (blank) frame, not in
place of it — the renderer stays mounted so a factory error still shows the
reason alongside the iframe (gating the frame off on error would drop the
auto-launched iframe an integration test relies on). Reworded the README row
and the test comment to say "below the frame"; added a code comment.
- #2 (applied): gate data-app-error on `running` (like data-app-status) so the
two attributes can't desync if a future path sets an error without running.
- #3 (no change): resource-read failures staying at data-app-status="loading"
(toast only) is intentional and documented.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BrihGWcrM9JGRyu41nzZYw
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(web): wire the listChanged indicator across all screens (#1402)
The per-screen list-changed indicator was hardcoded false in
InspectorView for Tools, Apps, Prompts, and Resources, so it could
never light up. Source it from the managed-state layer instead.
Each ManagedXState now tracks a listChanged flag: set true when its
*/list_changed notification arrives (the existing auto-refresh leaves
it set), exposed via getListChanged()/a listChangedChange event, and
cleared by clearListChanged() or on disconnect. The useManagedX hooks
surface the flag and clear it on a user-initiated refresh (the
indicator's Refresh button = acknowledge). App threads the three flags
to InspectorView, which maps them to the screens — Apps shares the
tools flag since apps are a filtered view of tools.
Tests cover set/clear/disconnect at the state-manager layer, the hook
subscription + refresh-clears behavior, and the InspectorView routing
(including Apps sharing the tools flag and no cross-wiring between
screens).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(core): close clear-after-await race in managed-list refresh (#1402 review)
The useManagedX refresh() callbacks cleared the listChanged flag AFTER
awaiting the fetch. If a */list_changed arrived mid-fetch, the state
re-set the flag (and auto-refreshed), then the trailing clear wiped that
genuinely-new signal. Clear BEFORE the await instead: the user's refresh
acknowledges only the change in hand, and a mid-fetch change re-sets the
flag and survives. Adds a race regression test on the tools hook.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(web): drop duplicate Refresh on the Apps screen (#1402)
AppControls uniquely had a standalone toolbar Refresh button AND the
ListChangedIndicator on a separate row, so once the indicator could
actually appear (this PR), the Apps screen showed two Refresh buttons.
Match the other three controls (Tools/Prompts/Resources): the indicator
is the single header-right refresh affordance. Updates AppControls and
AppsScreen tests to assert exactly one Refresh (and none when the list
hasn't changed).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(core): stop auto-invoking */list on list_changed; user pulls via Refresh (#1402)
With auto-refresh on, the list was already current by the time the
list-changed indicator appeared, making its Refresh button moot. Now the
managed Tools/Prompts/Resources state managers only flag the change on
*/list_changed (they no longer auto-refresh); the list stays as-is until
the user clicks Refresh, which fetches the new list and clears the
indicator. The connect-time initial load is unchanged.
Resource templates (no own indicator, part of the Resources screen) also
stop auto-refreshing; onRefreshResources now refreshes both resources and
templates, since a single resources/list_changed covers both. Tasks are
left auto-refreshing — they have no list-changed indicator and need live
status updates.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* docs(core): correct ManagedResourceTemplatesState header comments (#1402 review)
The class no longer subscribes to resourceTemplatesListChanged; update
the stale file/class docstrings that still claimed it syncs on that
notification to reflect connect + statusChange only.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(web): add per-server 'Auto Refresh on List Changed Notifications' option (#1402)
New Options section in ServerSettingsForm with a single checkbox. When
checked, the managed Tools/Prompts/Resources (and resource templates)
state auto-refreshes on */list_changed notifications; when unchecked
(default) the notification only lights the list-changed indicator and the
user pulls via Refresh.
Plumbing: new optional autoRefreshOnListChanged on InspectorServerSettings
and StoredMCPServer, round-tripped through serverList (lifted with a false
default, persisted only when true, added to the exhaustive field-key map);
surfaced from InspectorClient via a new getServerSettings() on the protocol;
the state managers read it live at notification time to decide auto-refresh
vs flag-only. Resource templates re-subscribe to their list_changed but only
auto-refresh when the option is on (they have no indicator of their own).
Tests cover the round-trip (true persists, false omitted), the form
section (render/checked/toggle), and each state manager's on/off branch.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(server): persist autoRefreshOnListChanged through validateSettings (#1402)
The PUT /api/servers write path runs settings through validateSettings,
which rebuilds the value from explicitly-named fields and silently drops
unknown keys — so autoRefreshOnListChanged never reached disk (the
serverList round-trip test passed because it calls
inspectorSettingsToStoredFields directly, bypassing this path). Validate
it as an optional boolean and include it in the built value; false is
omitted on disk by inspectorSettingsToStoredFields as before.
Adds integration tests over the real PUT path: true persists, false is
omitted, and a non-boolean is rejected (400).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(web): default the Server Settings modal to the Options section (#1402)
Open 'Options' instead of 'Custom Headers' when the Server Settings modal
first opens, so the new Auto Refresh toggle is the first thing surfaced.
Updates the modal default-expanded section, the modal test's initial-state
assertions, the three header tests (now expand 'Custom Headers' first since
it's no longer open by default), and the story defaults.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* chore: drop accidentally-committed Inspector history dump; gitignore the class
83076441 swept an exported Inspector message-history JSON (manual-testing
artifact at the repo root) into the tree via git add -A. Untrack it and
extend the existing repo-root debug-artifact .gitignore block (alongside
/*.png) to cover /inspector-history-*.json so these can't be re-added.
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): 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>
* 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>
* 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>
* 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>
cliffhall