No new inline comments this round, but two of the four suppressed ones were
real defects rather than restatements — both in code I added, both silently
wrong in the safe-looking direction.
**A symlinked secrets directory read as unmounted.** `isOnMountPoint` walks
to the nearest existing ancestor *lexically* and then string-prefixes it
against mountinfo, so `~/.mcp-inspector -> /data/inspector` with `/data`
mounted matched no mount point. The container then selected the session-only
store on a box that had explicitly arranged for persistence, and nothing said
so — the mount was there, the volume was there, the secrets evaporated on
exit. Resolved through symlinks now, tolerantly: a path that cannot be
resolved falls back to the lexical comparison, which is what this did before
and is correct whenever no symlink is involved. The `statSync` fallback branch
already followed links, so this also makes the two paths agree.
**The bulk read did not fix the case it was written for.** `getMany` batched
*one server's* fields, but both rehydration callers iterate servers — so a
catalog of 20 servers holding one secret each still paid 20 serialized scrypt
derivations. The same ~450ms stall the seam was introduced to remove, reached
one server at a time instead of one field at a time. Measuring the wrong unit
and then optimizing it is worse than not optimizing, because the number in the
commit message says it is fixed.
The seam is cross-server now — `getMany(requests: SecretBulkRequest[])` — and
both callers pass the whole catalog in one call. `FileSecretStore` decrypts
once for the entire rehydration; the keychain fallback still issues
independent parallel `get`s, which is what it wants. A test asserts one
derivation for four servers, next to the existing one asserting one for three
fields, so the property is pinned at both axes.
The other two suppressed comments are the cross-process guarantee — `mutate`'s
residual and the hand-off's TOCTOU delete — which remain decided against for
this PR and tracked in #2082.
Full `npm run ci` green.
Signed-off-by: cliffhall <cliff@futurescale.com>
Three inline findings and one of the two suppressed. The other suppressed
one — that the hand-off's delete is still a TOCTOU — is the cross-process
guarantee, decided against for this PR and tracked as #2082.
**My convergence tests did not converge anything.** The reviewer caught that
the "two stores racing" tests cannot race: `serialize` keys its queue on the
resolved path *process-wide*, which I made it do in the previous commit
precisely so two instances would take turns. So the tests passed by
serialization while their names and comments claimed they were exercising
the optimistic retry, and the retry's successful path had no coverage at all.
A fix I was pleased with silently disabled the tests written to prove it
worked.
Split in two now, both honest about what they drive:
- `in-process mutations are serialized per path` keeps the multi-instance
cases, which pin a real property — it is what makes the in-process case
correct *without* relying on the retry.
- `cross-process convergence` injects one external write between the store's
write and its verifying read, via raw `fs` so it bypasses the queue the way
another process would. The surviving entry can only be explained by the
re-apply, so these pin the mechanism rather than the outcome. Covers a
set, a delete, and the encrypted path where every attempt re-derives.
**A well-formed envelope is not an openable one.** `readOnDiskEncryption`
classified by structure alone, so a wrong or unset `MCP_INSPECTOR_SECRET_KEY`
produced the quiet, healthy "Encrypted file" footer for a file whose every
`get` returns `null` and every `set` is refused — the exact state the
unreadable footer promises to warn about. It now authenticates: decrypt and
discard, costing one scrypt derivation (~23ms) per descriptor build, once per
`/api/config`. A test asserted the old behavior and has been rewritten; the
comment on it explains why "describable" has to mean *accurately* describable.
**The copy button's `aria-label` was eating the warning.** `aria-label`
replaces descendant text, so a screen-reader user heard "Copy secrets file
path …" and never heard "Plaintext file" — the one sentence that should
change their mind about typing a secret. The status now leads the accessible
name and the path follows, which also keeps the name a superset of the
visible text (WCAG 2.5.3).
**The spec still described the pre-#1950 invariant** in its migration
bullets: every plaintext secret lifted into the keychain and removed from
`mcp.json`, with only `KeychainUnavailableError` named. Both bullets now
describe the selected store, the durability gate that withholds the disk
delete for a session-scoped store, the strict read behind the store-wins
lookup, and `SecretStoreUnavailableError` as the family the handler catches.
Full `npm run ci` green, including the per-file ≥90 gate — the new validator
and retry branches needed six more cases to clear it.
Signed-off-by: cliffhall <cliff@futurescale.com>
The headline is a correction to my own claim rather than a new defect.
**The optimistic verify is weaker than the comment said it was.** It
described the residual as a crash between the write and the verifying read.
The reviewer pointed out that no crash is needed: order two writers as
write-A, verify-A, write-B, verify-B and both report success while A's entry
is gone — A's verify simply ran before there was anything to see, and B did
nothing wrong. A crash is one instance of that shape, not the whole of it.
The trace in the old comment quietly assumed A verifies *after* B writes,
which nothing guarantees.
Corrected in the doc comment, the README and the spec, and the honest
summary is now stated rather than implied: this converges in every
interleaving where the clobber lands before the verify, it is not mutual
exclusion, and a deployment that needs the guarantee wants an OS-backed lock
from a dedicated library — not another hand-rolled election. It remains
better than what it replaced, which lost updates across a wider set of
interleavings and could not be closed at all with what Node exposes.
The other findings:
- **An envelope naming the cipher still is not an openable one.** Round 9
added a `kdf`/`data` presence check; three dot-separated strings satisfied
it while an empty tag, a 4-byte IV or `N: 3` still made every decrypt
throw — so the quiet "Encrypted file" footer went to a file whose next
save was guaranteed to fail. Now validates part sizes, a non-empty
ciphertext, and the scrypt cost parameters (node rejects a non-power-of-two
`N` synchronously, out of the derivation rather than the decrypt). A
positive control asserts a file we actually wrote still reads as encrypted.
- **`encryptionUnknown` + `looseMode` claimed an exposure it had not
established.** With both set, `plaintext` is absent, so a `!== false` test
landed on "anyone who can read it can read the secrets in it" — about a
file that may well hold ciphertext we simply could not classify. Three-way
now, in `secretStorageCaveat` and mirrored in the footer's tooltip, with a
test on each.
- **The README described a lock that no longer exists**, which is the worst
kind of stale doc: it stated a durability guarantee the code had stopped
providing. Rewritten to describe the actual design, including the residual.
- **The private-method access in the concurrency test** now carries the
justification the repo's TypeScript rules require: the asserted shape is
`writeMap`'s real signature, so a rename breaks the line instead of
silently detaching the stub, and a production injection seam would let the
real write path drift from the tested one.
Also updates `specification/v2_servers_file.md`, which still described #1356's
world: one keychain-only store, and unavailability as a hard error. It now
covers store selection and the probe, the file format and lazy encryption
upgrade, the concurrency design and its residual, the `getStrict`/`getMany`
seams and the two traps `getStrict` hit, the durability gate on migration,
the keychain hand-off and its conditional delete, and how the active store is
surfaced.
Full `npm run ci` green. (One unrelated `AppRenderer` theme-flip test failed
once under full-suite parallelism and passes in isolation with and without
these changes — a flake, not a regression.)
Signed-off-by: cliffhall <cliff@futurescale.com>
Two changes, interleaved in the same files because both are about the
same read-modify-write path.
## The lock is gone (maintainer decision on the round-5/7/8 finding)
Three consecutive review rounds found a real race in the `mkdir` lock, and
the last one is not patchable with what Node exposes: claiming a stale lock
needs compare-and-swap on a directory entry (`renameat2(RENAME_EXCHANGE)`).
Without it, `canSteal` inspects one lock but `rename` acts on whatever is at
that path *later*, so a waiter that lost the first race can move the
winner's fresh lock aside and enter alongside it.
Rather than patch it a fourth time, writers now collide and the loser
notices. `mutate` reads `M0`, applies the mutation to get `M1`, writes it,
then reads back `M2`; if they differ, someone wrote in between, so it
re-applies onto what they left and retries, bounded.
The comparison is over the **whole map**, and that is the load-bearing
detail. Checking only the entry we touched passes in exactly the case that
loses data — ours is present, and it is the *other* writer's that is gone.
With A and B from the same `M0`: A writes `MA`, B writes `MB` and clobbers
A; B verifies and correctly sees `MB`; A verifies, sees `M2 !== MA`,
re-applies onto `MB`, writes again. Both survive, and the writer that was
clobbered is the one that repairs it.
Two things this deliberately does not claim:
- It is **not mutual exclusion**. A process that crashes between its write
and its verifying read cannot notice it was clobbered, and nothing else
will. That window is strictly narrower than the lock's — which lost
updates with every participant alive — and needs no primitive Node lacks.
Stated in the code rather than left to be rediscovered in review.
- Non-convergence **fails loudly**. After five lost rounds `set` throws
rather than returning, because returning would reintroduce the silent
loss with extra steps. `delete` stays silent, per the interface contract.
`serialize` is now keyed on the resolved path **process-wide** rather than
per instance. Two `FileSecretStore`s on one file are ordinary — the resolved
store holds one and `absorbFileSecretsIntoKeyring` builds another — and a
per-instance queue does not order them against each other at all. The new
convergence test caught this on its first run: two instances racing `set`
lost an entry. In-process is now correct by construction; the verify covers
what it cannot see, a second Inspector process.
The hand-off's read → copy → delete was the lock's other user. Its delete is
now conditional: the file is re-read immediately before removal and compared
against what was copied, and it declines to delete when they differ. Leaving
the file costs nothing (keychain-wins makes the next run idempotent) while
deleting on a stale read costs a secret.
## Round 9 (two inline, two suppressed)
- **`KeyringSecretStore.getStrict` let raw binding errors escape.** Both
plaintext migrations keep their source file only when they catch
`SecretStoreUnavailableError`, so a transient keychain failure 500'd
`GET /api/servers` instead of abandoning the migration. Wrapped, as `set`
already does.
- **`FileSecretStore` had no strict read at all**, so `secretStoreGetStrict`
fell back to the tolerant `get` and mapped every read failure to `null` —
leaving the exact inversion the seam was introduced to prevent open on the
store it was introduced for.
- **`readOnDiskEncryption` reported "encrypted" for an envelope it cannot
open.** Naming the cipher is not the same as being openable: a file with no
`kdf`/`data`, or a `data` that is not `iv.tag.ciphertext`, produced the
quiet reassuring footer for a file whose next save was guaranteed to fail.
- **`/api/config` kept a stale descriptor when the resolver returned
`undefined`**, because the conditional spread left the startup value
standing. Assigned unconditionally now; `c.json` omits the key.
Full `npm run ci` green.
Signed-off-by: cliffhall <cliff@futurescale.com>
The lock finding from this round is deliberately not addressed here — see
below.
- **My round-6 strict-read fix was inert in production.** `DeferredSecretStore`
is what `defaultSecretStore()` returns, and it did not forward `getStrict`
— so `secretStoreGetStrict` fell through to the tolerant `get` for every
real caller, and the strictness existed only in tests that injected a
concrete store. Forwarded now, along with `getMany`.
- **The same tolerant-read bug remained in both pre-existing migrations.**
I added `getStrict` for the new hand-off in round 6 and never applied it to
the mcp.json and client.json migrations, which have the identical
shape: decide from `get`, then write on `null`. A transient keychain read
failure therefore let an older plaintext value overwrite a newer keychain
one — and the disk copy was then stripped, so the newer value was gone. Both
use the strict read now; the throw is a `SecretStoreUnavailableError`, which
both call sites already handle as "abandon the migration, keep the file".
- **The permission check made a false statement on Windows.** `stat` returns
synthetic bits there, and interpreting them yields a confident falsehood
either way — "0666 — not owner-only" about a file the ACL may restrict, or
"owner-only" about one it does not. The ACL is the real answer and this does
not inspect it, so Windows now reports `unknown`.
- **Rehydration cost one full decryption per field.** `get` reads and decrypts
the entire file, and both rehydration paths ask field by field, serialized
behind the store's own queue. Measured on this machine, scrypt at N=16384 is
~23ms per derivation, so a 20-secret encrypted catalog spent **~450ms of
pure key derivation on every `GET /api/servers`** — a visible stall, not a
micro-optimization. Added an optional `getMany` seam: `FileSecretStore`
decrypts once per server, everything else falls back to the parallel
per-field `get` that already suited it. A test counts derivations directly:
1 for a three-field `getMany`, 3 for the three `get`s it replaced.
Also broadens a log message that still said "Keychain unavailable" for a
store that may not be a keychain.
Full `npm run ci` green — 5,762 tests under the coverage gate.
Signed-off-by: cliffhall <cliff@futurescale.com>
No behavior change.
Round 7 removed the loose-mode `console.warn` from selection — it was
encryption-blind and duplicated the caveat printed moments later. This test
kept asserting `warn` was called with "0644" and kept passing, because the
caveat also contains the mode. So it was green while describing a code path
that no longer exists: named "warns once at selection", commented as if the
selection warning were the thing under test, and pinned to a substring loose
enough that either line would satisfy it.
Asserts the caveat's full sentence now, so it is fastened to the surviving
mechanism rather than to whichever line happens to mention the mode.
Also records what it took to reach this state at all, since it is not
obvious and I had it wrong: as the file's owner you *cannot* stage a
loose-mode descriptor, because `tightenSecretFilePermissions` repairs a 0644
file to 0600 during selection and the report is then correctly "owner-only".
It is only real for a file owned by another user or on a read-only mount,
which is why the chmod is mocked — standing in for a privilege a test cannot
drop.
Found by @mcp-inspector-v2-f1, who tried to screenshot the state for the PR
body and got "Owner-only permissions." instead.
Signed-off-by: cliffhall <cliff@futurescale.com>
Five taken; one rejected with evidence, below.
- **The keychain hand-off ran outside the file lock.** Read → copy → delete
has to be one transaction: another Inspector completing a `set` between
the read and the delete had its brand-new secret removed without ever
being copied — a write that reported success and then vanished, which is
the loss the lock exists to prevent, reached through the migration meant
to preserve things. The whole sequence now runs under the same
cross-process lock every mutation takes.
- **An unmigratable entry was skipped and the file deleted anyway.** A key
that is not `serverId:field` cannot be addressed through the store API,
so it cannot be copied — which makes the hand-off incomplete by
definition, and the complete-success rule then forbids removing the
source. It was discarding that value permanently in order to tidy up a
migration that had failed. The copyable entries still go across; the file
stays, and says why.
- **`readOnDiskEncryption` accepted any format version.** A version-2 file
was reported as a usable plaintext or encrypted store even though
`readMap` refuses it and every save fails. Checked before the mode now,
matching `readMap`.
- **The loose-mode warning at selection was encryption-blind**, claiming a
reader could read the secrets of a file that is ciphertext, and it
duplicated the caveat printed moments later by `warnAboutSecretStorage`.
Selection now only repairs; the descriptor path does the reporting,
because it is the one that knows whether the file is encrypted.
- **The footer's tooltip offered to copy the path without showing it.** The
band stopped printing it and the accessible name carries it for
screen-reader users, which left a sighted user unable to see where a
custom destination points without pasting it somewhere else.
**Not taken: "`writeStoreFile` is not atomic."** It is. The finding reads
`store-io.ts` line 84 as a direct write, but `writeFile` there is imported
from the `atomically` package, not `node:fs`. Verified rather than argued:
writing over an existing file changes its inode (so a rename happened) and
leaves mode `0600`. I had started implementing the suggested temp+rename
before checking the import, which would have replaced a tested library with
a hand-rolled duplicate.
That check corrected something of mine, though: `tightenSecretFilePermissions`
justified itself with "a rename over a 0644 file keeps the loose mode
forever", which is false — the temp carries `mode: 0o600` and the rename
replaces the inode, so every *write* re-establishes it. The function is
still needed for the case it actually addresses: a file that exists and is
never written. Comment corrected to say so.
Full `npm run ci` green — 5,753 tests under the coverage gate.
Signed-off-by: cliffhall <cliff@futurescale.com>
Four findings, all suppressed, all real. Three are the same mistake in
different places: a state that was never established being reported as a
confident answer.
- **`readOnDiskEncryption` conflated "no file" with "unreadable file".**
Both answered `null`, and the descriptor falls back to the configured
write policy on `null` — so a passphrase-configured install reported
"File (encrypted)" about a file whose envelope had never parsed, and
which `set` was in fact about to refuse rather than overwrite. It now
returns a four-state result, and only `absent` falls back to the policy
(honestly: the first write really will create it that way). `unreadable`
produces `encryptionUnknown` on the descriptor, `plaintext` is omitted
entirely so nothing can read a default out of it, and the label, caveat,
tone and footer all carry "unreadable" rather than guessing a mode.
- **The loose-mode warning overstated exposure for an encrypted file.**
"Anyone who can read it can read the secrets in it" is false when the
file is ciphertext — a reader gets the bytes and still needs the
passphrase. Both the caveat and the footer tooltip now branch on
encryption: the passphrase becomes the only protection, which is the
real consequence. Overstating it is not harmless, because a warning that
cries wolf is discounted on the occasion it is literally true.
- **The keychain hand-off could invert its own keychain-wins rule.**
`SecretStore.get` answers `null` both for a missing entry and for a
keychain it could not read, and the migration *writes* on `null` — so a
transient read failure would overwrite a newer keychain value with the
older file copy. Added an optional `getStrict` (throws instead of
swallowing) implemented by `KeyringSecretStore`, with
`secretStoreGetStrict` falling back to `get` for stores whose reads
cannot fail. A throw now aborts the hand-off with the file left in
place, so the next run retries.
Writing the tests turned up one more, in code the review had passed:
`parseOwner` recorded a garbled pid as `-1`, and `process.kill(-1, 0)` is
a *wildcard* asking about every process this user can signal — it answers
"alive" almost always, which would have made a lock with a corrupt stamp
permanently unstealable. Only a positive pid is usable now; anything else
is treated as no pid at all and the mtime decides.
Tests: the four envelope states kept distinct; the descriptor refusing to
guess, with the genuine no-file case as the control; strict-read abort
without a write and with the file preserved, plus a store whose `get` and
`getStrict` disagree so the test proves which one the migration consults;
encryption-aware wording asserted in both directions in the caveat and the
band; three unusable owner stamps and a wildcard pid; and the descriptor
shapes that must now be rejected — neither half of the encryption question
answered, or both. Full `npm run ci` green — 5,750 tests under the
coverage gate.
Signed-off-by: cliffhall <cliff@futurescale.com>
Repairs origin as well as answering the review. 2d2c787e committed the
`absorbFileSecretsIntoKeyring` tests without the implementation they
import, so the branch has been failing `tsc` since; this commit brings the
source across.
Round 5's four findings (two inline, two suppressed):
- **The stale-lock takeover was still racy.** Round 4 made *stealing*
liveness-aware but left the takeover as remove-then-recreate, which is
not an election: two waiters can both judge the same lock stale, the
first removes it and takes a fresh one, and the second then removes
*that* — both inside the section, which is the lost update the takeover
exists to prevent. Claiming is now an atomic `rename` to a
token-unique name, so exactly one waiter can take a given stale lock and
the losers get ENOENT. A failed claim also falls through to the deadline
instead of `continue`ing, so a takeover that can never succeed times out
rather than spinning forever on a swallowed error.
- **Every `stat` failure was read as "no file yet".** An existing file we
cannot inspect (EACCES on the directory, a mode-less filesystem) then
produced no warning at all, and the footer went on stating owner-only
permissions having verified nothing. There is now a third state,
`unknown`, carried end to end: `permissionsUnknown` on the descriptor, a
caveat that says the permissions could not be checked, a `warn` tone, and
its own branch in the footer sentence — which otherwise would have
claimed "Owner-only permissions." by omission, the same bug one layer up.
- **mountinfo decoded only `\040`.** The kernel escapes tab, newline and
backslash too (`\011`, `\012`, `\134`), so a mounted path containing any
of them was compared in its escaped form, matched nothing, and read as
the container's writable layer — `memory` selected and persistence
silently lost on a box that had arranged for it. Decoded as octal in
general now, since the rule is octal-in-general and a future addition
would otherwise reintroduce exactly this.
- **File-backed secrets vanished once a keychain appeared.** Save secrets
without libsecret, install libsecret, and the next run selects the
keychain and stops seeing them: still on disk, read by nothing, and
nothing looks broken. `absorbFileSecretsIntoKeyring` hands them over on
the run that finds a keychain, under the rules the two existing plaintext
migrations already use — keychain wins on conflict, the file is removed
only on complete success, and a file that cannot be read is never
deleted, because that is precisely the case where the values would be
lost.
Tests: a two-waiter race over one stale lock asserting a single holder
throughout; `unknown` from `readSecretFilePermissions` through the
descriptor to the footer sentence and tone; the hand-off's happy path,
keychain-wins, write-failure, undecryptable-file, empty-file,
malformed-key, and explicitly-configured-keyring cases; and six more
rejected descriptor shapes, including each file-only field on a non-file
kind. Full `npm run ci` green — 5,732 tests under the coverage gate.
Signed-off-by: cliffhall <cliff@futurescale.com>
Reads "Memory (this session only): Secrets are not written anywhere and
are lost on exit." — shorter tail than "lost when the Inspector exits",
and a colon separating the store from what follows it.
The colon is added at the footer's composition, not inside
`secretStorageLabel`. The label is also consumed alone by
`secretStorageSummary`, which appends " at <path>" for a file-backed
store, so a colon baked into the label would print
"File (encrypted): at /path" in the startup banner. It renders only when
a detail actually follows.
The caveat string is shared with the banner's warning line, so the
terminal picks up the shorter wording too.
Signed-off-by: cliffhall <cliff@futurescale.com>
Found on a re-read of the round-4 liveness work, not by review.
The heartbeat refreshed the lock directory's mtime unconditionally. After
a steal — possible when this process looked dead, or across hosts where
the pid check cannot apply — the directory belongs to someone else, and
touching it keeps *their* lock looking alive. If they then die, the stale
timestamp nobody is maintaining is being maintained by us, and a third
writer can never break in: a stale lock turned unbreakable, which is the
permanent outage the stealing exists to avoid.
The heartbeat now re-reads the owner stamp and clears itself when the
token no longer matches, so it only ever refreshes a lock we still hold.
The release path already checked the token; this closes the same hole on
the other side.
Tested by stamping a different owner onto a held lock and aging it: the
mtime stays old across several heartbeat intervals, and release leaves the
thief's lock intact.
Signed-off-by: cliffhall <cliff@futurescale.com>
Four findings (two inline, two suppressed), all taken. Three are the lock,
which is where the remaining sharp edges were.
- **A lock could be created and then never stamped.** If `mkdir` succeeded
but writing the owner token failed, the catch treated it as "cannot lock
here", returned the no-op release, and left the directory behind. The
mutation then ran unlocked *and* every later writer queued behind an
orphan nobody could deliberately release, until it aged out. The
acquisition now tracks whether it created the directory, removes it on a
failed stamp, and refuses rather than entering the section unlocked.
- **An old mtime was being treated as proof the owner was dead.** It is
not: a holder that is SIGSTOPed, starved by a blocked event loop, or on a
machine that suspended is alive and not running, and expiring it on
elapsed time alone takes the lock out from under a live critical section
— the lost update the lock exists to prevent, arrived at through the
machinery meant to prevent it. The heartbeat narrowed the window; it
could not close it, because the holder that most needs protecting is
precisely the one not running its timers.
The owner stamp now carries `pid` and `host`, and staleness only *opens*
the question — on the same host a waiter asks the OS whether that pid
still exists (`kill(pid, 0)`, which sends no signal; `EPERM` counts as
alive, so a holder running as another user is never stolen from). Only a
genuinely absent process is stolen from. Where the stamp cannot be
trusted — another machine over a shared filesystem, or a missing/corrupt
owner file — the mtime remains the answer, because the alternative is a
lock nothing can ever break, turning one dead process into a permanent
outage.
- **Podman was classified as a host.** `/.dockerenv` is Docker's marker and
rootless Podman has no equivalent of it, while its cgroup paths say
`libpod-<id>.scope` rather than the literal "podman" the pattern looked
for. Such a box fell back to a file in the ephemeral writable layer and
reported it as durable — the wrong way round to be wrong.
`/run/.containerenv` and `libpod` are both recognized now.
- **Two tests floated a second `acquireLock`**, so the test could finish and
`afterEach` could remove the temp directory while the acquisition was
still running. Both hold and settle it now.
Separately, a defect of my own found while re-reading round 3:
`describeFileStore` called `tightenSecretFilePermissions`, so the function
whose job is to *describe* the store issued a `chmod` on every
`GET /api/config` — once per page load, on a file another process may be
mid-write on. Split out a read-only `readSecretFilePermissions`; the repair
stays at startup, where it happens once and its outcome is announced.
Tests: a stale lock whose owner is alive is not stolen (and survives the
attempt), one whose pid is gone is, one stamped by another host is; a
failed ownership stamp reports and leaves nothing behind; Podman via both
signals; and the read-only permission check leaves a 0644 file at 0644.
Full `npm run ci` green — 5,711 tests under the coverage gate.
Signed-off-by: cliffhall <cliff@futurescale.com>
Nine findings (six inline, three suppressed), all taken. Two were serious.
**The memory fallback made the migrations destructive.** Both migrations —
mcp.json in the remote server, client.json in `node-persistence` — lift a
plaintext secret off disk, write it to the secret store, and then delete it
from the file. That trade is safe only while the store is at least as
durable as the disk, which held by construction while `InMemorySecretStore`
was a test double. Making it a production fallback broke it: on an
unmounted container a plain `GET /api/servers` would have moved a user's
existing secrets into RAM and lost them at exit — destroying data on a
read.
`SecretStore` gains an optional `isDurable()`, absent meaning durable, and
both migrations withhold the disk delete when it answers false (the value
is still loaded, so the session behaves normally). The production fallback
is now a named `SessionSecretStore` rather than the bare
`InMemorySecretStore`, so the suite's many doubles keep standing in for a
*working keychain* — which is what they are there to be — and only code
that deliberately chooses RAM in production says so.
**Round 2's per-call descriptor never reached anyone.** Both web servers
resolved it once at startup and baked it into `initialConfig`, which
`/api/config` serves unchanged — so the freshness fix was inert for its
only consumer. The route now takes a `secretStorageResolver` and calls it
per request.
The lock needed three more fixes, and they compound:
- **Release removed whatever directory was at the path.** A holder whose
work outran the stale threshold would delete the *next* owner's lock,
letting a third writer into a section two processes believed they held.
Ownership is now stamped with a random token and release only removes a
lock still carrying it.
- **The timeout fell through to an unlocked write** — under contention,
the one moment another writer is provably active. It now throws
`SecretFileLockTimeoutError`, which `set` is allowed to do.
- **The first write of every process bypassed the lock**, because the
secrets directory does not exist yet and `mkdir` failed ENOENT. The
parent is created first. First writes are exactly when two processes are
most likely to start together.
Fixing the timeout exposed a fourth: with stale (10s) below timeout, a
waiter stole from a holder that was merely *slow* rather than dead —
reinstating the lost update. A holder now heartbeats its lock, so
"stale" means the holder is gone, and the timeout is what a live-but-slow
holder produces. Lock timings are injectable so the tests can exercise
contention, staleness, and timeout without spending the real intervals.
Also: the loose-permission state is carried into `SecretStorageInfo` as
`looseMode` and rendered as the caveat (it outranks the encryption caveat —
a file others can read is a live exposure, while "unencrypted" is a
property of a file only its owner can open), rather than living in a
startup log the browser cannot see; and `usableSecretStorage` validates
`pendingEncryption` and `looseMode` as the optional booleans/numbers they
are (`pendingEncryption: "false"` is truthy and would have printed advice
that is the opposite of the truth) and rejects file-only fields on a
keychain or memory descriptor.
Tests: both migrations against a session store with a durable control
alongside, so the guard is visibly the difference; `/api/config` serving a
changing resolver and falling back without one; lock ownership after a
steal, the timeout under a live holder, no-steal from a slow holder, and a
locked first write; `looseMode` through the descriptor, the caveat
ordering, and the tone; and five more rejected payload shapes. Full
`npm run ci` green — 5,704 tests under the coverage gate.
Signed-off-by: cliffhall <cliff@futurescale.com>
Seven findings, all taken.
- **The descriptor went stale after the upgrading write.** Round 1 made it
read the file instead of the policy, but it was still computed once and
cached beside the selection — so the first `set` under a newly-set
passphrase encrypted the file while `/api/config` kept serving "still
unencrypted" for the rest of the session, describing bytes this very
process had just changed. The *selection* stays cached (the banner, the
payload, and the store doing the writing must agree on which store it
is); the file descriptor is re-derived per call, which costs one read of
a small file per config fetch.
- **The mount check asked the wrong question.** Comparing a directory's
`st_dev` against its parent's detects a mount *at* that path, but a
volume mounted at `/home/node` leaves `/home/node/.mcp-inspector` an
ordinary subdirectory sharing its parent's device — so a durable
container setup was judged unmounted and demoted to session-only
`memory`. It now reads `/proc/self/mountinfo` and asks whether the
directory is *under* any mount other than `/`, which also catches the
same-device bind mount no `st_dev` test can see. Where mountinfo is
unreadable (macOS, Windows, a `/proc`-less container) it falls back to
the device comparison, now against `/` rather than the parent.
- **A decrypted payload was cast, not checked.** GCM proves the bytes are
authentic, not that they parse to the right shape. An array passes
`typeof === "object"`, takes the named assignment, and then
`JSON.stringify` drops every named property of an array — so `set` would
resolve having written a file without the secret it was handed. Silent
loss on the one operation whose entire contract is that it does not lose
things. Both branches now validate a non-array object of string values,
and a shape refusal is deliberately not rethrown as a key mismatch,
which would send the user to fix a working passphrase.
- **The read-modify-write cycle was only serialized in-process.** A
durable file is shared state and a second Inspector on the same box (a
CLI run beside a web session) could read the same map and overwrite the
first one's save, both reporting success. Mutations now take a
`mkdir`-based lock. A lock left by a killed process is stolen after 10s
— one `docker kill` must not brick secret writes permanently, which
would be a worse failure than the race. If the lock cannot be created at
all, the write proceeds unlocked rather than denying a real user a real
secret to prevent a race with a process that may not exist. Reads do not
lock: `writeStoreFile` is atomic, so a reader sees the old file or the
new one, never a torn one.
- **The permission repair was fire-and-forget while the docs promised
0600.** A file owned by another user, or on a read-only mount, stayed
exposed while the footer described it as protected. It now verifies the
resulting mode and warns, naming the actual mode, when it could not be
tightened.
- **`Group`'s theme carried a domain-named `secretStorageFooter`
variant**, which AGENTS.md reserves for app-wide primitive styling.
Renamed to the structural `stickyModalFooter`; the secret-storage
meaning stays in the element that owns it.
- The PR description still claimed the salt made a short passphrase safe,
contradicting the corrected comment and README. Updated.
Tests: both mount layouts that motivated the change (mount-at and
mount-above) plus the sibling-prefix, space-escaped, first-run, and
no-mountinfo cases; the array and non-string refusals on both the
plaintext and the authentic-ciphertext paths (the latter through a
hand-built fixture, since the store's own writer cannot produce that
state); lock exclusion, stale-lock stealing, the un-lockable fallback, and
two independent store instances racing; the permission report in all three
states; and the descriptor no longer going stale within one process. Full
`npm run ci` green — 5,688 tests under the coverage gate.
Signed-off-by: cliffhall <cliff@futurescale.com>
Six findings, all taken.
- The descriptor reported the *write policy*, not the file. Adding
`MCP_INSPECTOR_SECRET_KEY` to an install that already has a plaintext
file flipped `store.encrypted` immediately, so the banner and footer
claimed "File (encrypted)" while the existing secrets sat readable
until the next `set`. `FileSecretStore.readOnDiskEncryption()` now
reads the envelope (never the payload, so it needs no passphrase) and
the descriptor is built from that. The transitional state gets its own
`pendingEncryption` flag, which changes the advice without softening
the verdict — "set a passphrase" is the wrong thing to tell someone who
already has, so the caveat names the pending write instead.
- The secrets file ignored the configured storage directory. A container
mounting only that directory was judged unmounted, fell back to
`memory`, and lost secrets the user had arranged to keep.
`defaultSecretFilePath()` now follows `MCP_STORAGE_DIR` — the variable
that already relocates OAuth tokens and `client.json` — between the
explicit `MCP_INSPECTOR_SECRET_FILE` and the `~/.mcp-inspector`
default. (The issue comment called it `MCP_INSPECTOR_STORAGE_DIR`; no
such variable exists.)
- `version` was never checked before the plaintext branch, so a
`{ version: 2, encryption: "none" }` file was read and silently
rewritten as version 1, discarding whatever the newer format added —
the same destroy-to-satisfy-an-additive-request case the encrypted
branch already refuses. Checked first now, so the answer no longer
depends on which encryption mode the newer writer chose.
- The doc comment claimed a per-file salt made "a short, memorable value
safe to use". It does not: the salt defeats precomputation, not
guessing, and the scrypt cost is deliberately a few milliseconds
because it runs on every read and write. Both the comment and the
README now say to use a high-entropy value, generated rather than
chosen.
- `usableSecretStorage` validated only `kind`, so a partial
`{ kind: "file" }` was cast through and — because a missing `plaintext`
is falsy — rendered as the quiet, neutral *encrypted* file. That is the
confident half-answer the guard exists to reject, produced by the
absence of information rather than by any claim the backend made.
`reason`, `durable`, and each kind's own required fields are checked.
- The footer's inline `Tooltip` carried four static props; extracted to a
`CaveatTooltip` constant per the `.withProps()` rule.
Tests cover each: the policy/file disagreement in both directions and the
no-file-yet fallback, `MCP_STORAGE_DIR` and its precedence against
`MCP_INSPECTOR_SECRET_FILE`, the version refusal (asserting the bytes
survive it) split from the unrecognized-cipher refusal it used to be
conflated with, `readOnDiskEncryption` including the keyless and
unparseable cases, five new rejected descriptor shapes, and the changed
caveat. Full `npm run ci` green.
Signed-off-by: cliffhall <cliff@futurescale.com>
Closes#1950
Before this, a box with no reachable OS keychain — the published
container (no D-Bus session, #1848), Android/Termux (no platform binary,
#1905), a minimal Linux install without libsecret — could not persist an
OAuth client secret or a stdio `env:` value at all. `KeyringSecretStore`
degraded into a store that silently holds nothing: `get` returned null,
`delete` no-opped, and `set` hard-failed with a 503. That is the correct
failure behavior, but it left the documented, supported way to run the
Inspector with no secret persistence whatsoever.
Adds the two missing implementations and, more importantly, the policy
that picks one and the surfaces that say which was picked.
- `FileSecretStore` — one `0600` JSON document at
`~/.mcp-inspector/secrets.json`, encrypted with AES-256-GCM whenever
`MCP_INSPECTOR_SECRET_KEY` is set (scrypt, per-file random salt). The
whole map is encrypted as a unit rather than value-by-value so the
account names — `${serverId}:${field}` — are hidden too; a per-value
scheme would leave a readable index of which servers you hold a client
secret for. A file that can no longer be decrypted reads as empty
(matching the interface's read tolerance) but *refuses to be written*,
because replacing a file of still-valid secrets to satisfy an additive
request destroys data.
- `secret-store-selection.ts` — the policy.
`MCP_INSPECTOR_SECRET_STORE=keyring|file|memory` wins outright;
otherwise probe the keychain and, if unreachable, fall back loudly. The
fallback is `memory` in a container whose secrets directory is not on a
mount, and `file` otherwise — so mounting the volume the README already
recommends for the catalog flips the same run to durable storage with
no configuration, and an unmounted container gets the honest answer
rather than a file `--rm` will discard. Cached per process so the
banner, `/api/config`, and the store doing the writing cannot disagree.
- Automatic, but never silent. The downgrade prints at startup, rides
`GET /api/config`, and is named in a permanent 32px footer at the
bottom of both dialogs that accept a secret. A startup banner is seen
once by whoever started the process; a toast is seen once; a
dismissible banner is the thing a user dismisses before doing the work
it describes. The plaintext and memory cases render their caveat inline
in a warning tone — a warning you must hover to discover is not one.
- `SecretStoreUnavailableError` is now the base every store's `set`
throws, so the routes translate one condition into a 503 instead of
matching `KeychainUnavailableError` by name.
Docs: the container section of the root README gains a table of which
store you get and how to make secrets survive, plus the
`MCP_INSPECTOR_SECRET_KEY` upgrade/mismatch rules.
Tests: the existing interface contract block is extracted to
`secretStoreContract.ts` and run against all three implementations, so
none can drift on the availability contract. `FileSecretStore` is tested
against a real temp filesystem — mode, cross-instance round-trip, that
the ciphertext does not contain the plaintext, the read-modify-write
race, and every refusal path. Full `npm run ci` passes, per-file gate
included.
Signed-off-by: cliffhall <cliff@futurescale.com>
Copilot review round 1. The helper's contract is exactly-once teardown
and the test asserted it must not lean on Hono's own re-entry guard, but
neither the implementation nor the test actually enforced that: the
listener had no local guard and the test aborted only once.
Signed-off-by: cliffhall <cliff@futurescale.com>
Both SSE handlers installed their subscriber, primed the stream so
Firefox's fetch() resolves (#1858), and only then registered onAbort.
Hono's onAbort has no already-aborted replay and abort() fires only the
subscribers registered at that moment, so a client disconnecting while
the priming write was in flight fired nothing: the session event
consumer / server-list subscriber stayed installed, the mcp.json watcher
was never stopped, and the handler parked forever on a promise nothing
would resolve.
Add primeAndHoldSseStream, which registers one listener before the first
await — it both cleans up and releases the hold, collapsing the previous
duplicate onAbort registrations. Priming still happens after the
subscriber is installed, since callers treat the stream's first bytes as
proof they are subscribed.
Closes#1999
Signed-off-by: cliffhall <cliff@futurescale.com>
Addresses Copilot's round-15 review on #2035.
1. `readResourceFromTemplate` had no regression distinguishing this module
from the SDK expander it replaced: the existing test uses `report.txt`
and the transport tests expand separately and call `readResource`, so all
of them pass either way.
Writing that test corrected something I had asserted twice in this PR.
The first version used `foo/bar` -- and verified by reverting the wiring,
it still passed. `encodeURIComponent` DOES encode `/`, so the SDK expands
`foobar://events/{topic}` to `foobar://events/foo%2Fbar` exactly as this
module does. #1919 was the WEB PANEL's own `String.replace`, not the SDK
expander, which the TUI path was using correctly all along.
The test now uses `a!b`. `!` is a sub-delim: RFC 6570 requires it
encoded, `encodeURIComponent` leaves it bare, so the SDK returns
`foobar://events/a!b` and the assertion fails. Re-verified by reverting
the wiring -- 2 of the 10 tests fail (this one and the malformed-template
refusal), where before the change none did.
2. The module summary said "shared by every client" a dozen lines above the
paragraph explaining the CLI is not a consumer. Now says web and TUI.
Also records, on MAX_VALUE_LENGTH, the audit prompted by round 14: the
per-value ceiling was the ONLY guard lost with `UriTemplate.expand`. The
template-length (1e6), expression-count (1e4) and variable-name-length (1e6)
limits all live in the SDK's `parse()`, which runs from the constructor --
and strict expansion still constructs a `UriTemplate`, so they remain
enforced. The restored limit is the SDK's `MAX_VARIABLE_LENGTH`, read from
the pinned dist rather than assumed.
npm run ci passes.
Signed-off-by: cliffhall <cliff@futurescale.com>
Addresses Copilot's round-8 review on #2035 (one inline comment, one
suppressed). Both real.
1. RFC 6570 distinguishes an UNDEFINED variable, which omits its whole
expression, from one defined as the empty string, which expands. The
expander was calling `definedValues` on the way in, collapsing the two --
so `x{?q}` with `{ q: "" }` gave `x` instead of `x?q=`, and there was no
way for any caller, `readResourceFromTemplate` included, to request
either that or `x;q`. Key presence is now honored, including §3.2.7's
rule that the matrix operator drops the `=` for an empty value (`x{;q}`
-> `x;q`, while `?`/`&` keep it).
Dropping blanks is a FORM concern, so it moved to the two form
boundaries: the web panel seeds every declared variable with "" and the
TUI's ink-form returns "" for every untouched field, so neither can tell
"never touched" from "deliberately empty" -- but that is a fact about a
text input, not about the template, and it has no business deciding what
a template means for every other caller.
Six unit tests and one integration test encoded the old collapse and now
assert the corrected meaning. The prototype-collision cases get stronger
rather than weaker: a blank `{?toString}` expands to `?toString=` -- its
OWN empty value -- where the original bug expanded the inherited
function's body.
2. The expansion error is now `role="alert"`. It can appear while typing (a
pasted value that cannot be encoded), and plain text is not announced, so
a screen-reader user met a disabled Read Resource with no stated reason.
npm run ci passes.
Signed-off-by: cliffhall <cliff@futurescale.com>
Addresses Copilot's round-7 review on #2035. Two suppressed comments, both
real; each reproduced against this branch before acting.
1. `parseVarSpec` was a "strip the first `*`, trim, split on `:`" pass, so
it accepted shapes RFC 6570's varspec production does not admit --
measured on the previous commit:
x://{*id} name "id" => "x://abcdef" explode is a TRAILING modifier
x://{id*:3} name "id" => "x://abc" the two modifiers are exclusive
x://{ id } name "id" => "x://abcdef" whitespace is not a varchar
x://{a b} name "a b" => "x://1" ditto -- and it emits into the URI
`{id*:3}` is the worst of them: it silently truncated the value through a
combination the grammar forbids. Parsing is now one anchored regex --
a varname, then at most one trailing `*` or `:1-9999`.
The name set is `varchar` widened by `-` and `~`, deliberately. A name is
the one thing expansion emits into the URI WITHOUT encoding it (`;`, `?`
and `&` write `name=value`), so the rule it has to satisfy is "needs no
encoding" -- RFC 3986 unreserved plus pct-encoded, of which the RFC's own
varchar is a subset. Strict varchar would refuse `{user-id}`, which real
servers publish and which the SDK's matcher round-trips; refusing to read
a resource that demonstrably works is a worse outcome for a debugging
tool than a name one character outside the production. Everything now
rejected is a shape we would otherwise have to guess at.
2. The TUI built its "Missing required template variable(s): ..." list with
its own bare `values[name]` filter while gating on `hasRequiredValues`,
which reads own properties only. For a variable legitimately named
`constructor` the two disagreed: `({})["constructor"]` is `Object`, whose
`.length` is 1, so the list came out empty and the error named no field
at all while the submit stayed blocked. Both now come from one exported
`unmetRequiredGroups`, with `hasRequiredValues` defined in terms of it.
Also renames a test whose title still said multi-name query expressions were
left "to the SDK, which handles them" -- expansion has been owned here since
935113d1; the behavior it asserts is naming, not delegation.
npm run ci passes.
Signed-off-by: cliffhall <cliff@futurescale.com>
Copilot round 2 (PR #2039) surfaced no new visible comments but six
suppressed ones. Four were actionable.
`clients/tui/src/App.tsx` — a stale-completion race in the new banner:
start disconnecting server A, switch to B, then let A's disconnect
reject, and A's failure appears under B. Rapid same-server retries had
the same shape, and the stale attempt is the one most likely to reject.
Claim a monotonic attempt token before awaiting and publish only if it
is still current and the server has not changed; `handleConnect` bumps
the same counter so a connect retires an in-flight disconnect. The
rejection is still owned when superseded — just not published. The
regression test drives a deferred rejection across a server switch, and
fails without the guard.
`tsconfig.lint.json` — the root type-aware block matches
`test-servers/src/**/*.tsx` but the parser project omitted that
extension, so such a file would have failed with "not found in any of
the provided project(s)" instead of being linted. Added, and AGENTS.md
now states the invariant: the include must stay a superset of the lint
globs.
`clients/web/server/{server,vite-hono-plugin}.ts` — the `open()`
rejection branch was untested, and duplicated. Both were fixed by
removing the duplication: `openBrowser` now lives beside
`printServerBanner` in `web-server-config.ts`, with one implementation
to test. Two copies of a one-line `.catch` is two chances for one to
drift back to a bare `open(url)`. The dev plugin is not booted in tests
— it returns early under VITEST by design — but it calls the same
tested helper.
`core/mcp/remote/node/server.ts` — `closeAbortedStream`'s rejecting-close
path, the entire reason the helper exists, had no test. Exported it (as
this file already does for `requestIdForSendWait` / `mcpParamHeadersOnly`)
and covered both legs, asserting no unhandledRejection fires.
Also corrects AGENTS.md, which said each client lists "both" tsconfig
projects — true for cli/tui/launcher, but web lists four leaf projects.
Signed-off-by: cliffhall <cliff@futurescale.com>
Addresses Copilot's round-5 review on #2035 — eight suppressed comments,
all of them real.
The consequential one overturns behavior I copied from the SDK and
defended in two earlier rounds. RFC 6570 expands each expression
independently; it carries no cross-expression state. The SDK tracks
whether a query expression has already emitted and rewrites a later
`{?two}`'s `?` to `&`. That looks friendlier and is wrong, and its own
matcher says so: for `x{?one}{?two}` it expands to `x?one=1&two=2`, and
`match("x?one=1&two=2")` on that same template returns **null**, while
`match("x?one=1?two=2")` returns both variables. So the rewrite emitted a
URI the advertised template cannot match — #1919 one level up. A server
that wants a continuation advertises `{?one}{&two}`, which expands to `&`
and matches; producing that shape is the server's choice to declare, not
ours to infer.
Also:
- `{id:abc}` was silently treated as `{id}`, and a test codified it. RFC
6570's max-length is `%x31-39 0*3DIGIT`, so `{id:}`, `{id:0}`,
`{id:abc}` and `{id:10000}` are invalid templates rather than templates
with an ignorable modifier — and the SDK's constructor accepts them all,
so nothing else rejects them. Strict expansion now throws; lenient
returns the raw template. Discovery stays lenient so the panel still
renders rather than going blank.
- The TUI could submit `{a,b}` completely blank while the web panel
blocked the same request: ink-form has no way to express "any one of
these", so its members are left optional and the modal needs its own
group check. Added, with tests — it was also below the coverage gate
without them (functions 76.92%).
- Added the integration test the config never had: it resolves the
checked-in config and drives all four reads over a real transport, so a
misspelt preset fails there rather than only when someone runs the repro
by hand.
- "All three clients" was false. The CLI has no template form and its
`resources/read` passes the already-expanded `--uri` straight to
`readResource`, so nothing here runs for it. Corrected in the README,
AGENTS.md, and the module comment, and dropped a docblock still
describing the SDK-delegation path removed last round.
Signed-off-by: cliffhall <cliff@futurescale.com>
Adds two optional per-server fields — `oauth.authorizationUrl` and
`oauth.tokenUrl` — that override the endpoints authorization-server
metadata discovery resolved, so a server can be pointed at a development
or staging authorization server without changing what it advertises.
SDK v2 routes both endpoints through the discovered metadata document and
neither through OAuthClientProvider, so the override is applied by
wrapping the client's base fetch and patching the metadata document in
flight (core/auth/endpointOverrides.ts). Wrapping the base fetch — rather
than only the auth fetch — also covers the discovery the SDK runs from
inside the transport on the 401/refresh path.
The fields round-trip through mcp.json, the /api/servers routes, the
runner (CLI/TUI), and the Server Settings → Authorization panel, where a
value that is not an absolute http(s) URL is flagged inline; at runtime
such a value is dropped with a single warning rather than failing the
connection.
Signed-off-by: cliffhall <cliff@futurescale.com>
Copilot review: the unit test covered the panel's rendering, but nothing
executed `mrtr_empty`, so the handler, its preset-registry branch, and the
showcase config entry could regress with CI green.
Adds a live modern-era test that loads `mrtr-showcase-http.json` through
`loadConfig`/`resolveConfig` — the same artifacts the README's manual
reproduction uses, so the config entry and the registry branch are covered
rather than just the fixture factory — answers the embedded elicitation, and
asserts the terminal result really is an empty `CallToolResult`: that is the
input the Tools panel must render as "completed, returned nothing".
Signed-off-by: cliffhall <cliff@futurescale.com>
Copilot round 2 (suppressed comments). Five of the seven findings were the
accessibility of the new row list and one stray cast; all are addressed here.
- The reserved-key reason is now the key input's `error` **string** rather than
a bare boolean plus free-standing text. Mantine renders it in the input's own
error slot and wires the `aria-describedby` association, so a screen reader
learns *why* the field is invalid; a boolean only set `aria-invalid`. This
also deletes the duplicate `ReservedParamError` text and its Stack wrapper.
- Each control carries a per-row `aria-label` — the section heading and the
placeholders are not programmatically associated with the inputs, so the key
box, the value box, and a remove button announcing only "X" were
indistinguishable to assistive technology. The label names the row by its key,
falling back to "row N" while the key is still blank.
- Dropped an unnecessary `as unknown as Record<string, unknown>` in the
servers-route integration test: `MCPConfig.mcpServers` is already typed as
`Record<string, StoredMCPServer>`, so the entry's `oauth` reads directly.
Three tests cover the new behavior (accessible names, the blank-key row-number
fallback, and the aria-describedby association), each verified to fail against
the pre-fix component. No change to the reserved-key rule or the EMA boundary.
Signed-off-by: cliffhall <cliff@futurescale.com>
Copilot round 10.
The scan no longer publishes its malformed entries into the tools/list report
at all. Round 9 only stopped it overwriting the aggregate's; the remaining
case is unsalvageable for a stronger reason than 'two different requests':
- a FAILED strict aggregate runs salvage, which writes the report itself;
- a SUCCESSFUL one means every entry parsed, so the rendered list contains no
malformed entries.
So by the time the scan finds something, either the aggregate already reported
it with indices matching the screen, or the aggregate succeeded and what the
scan found is provably not in the list being displayed. Publishing could only
ever name entries the user cannot see.
Nothing is lost: listToolsForScan still marks its own refused response, so the
finding surfaces on the Protocol entry for the exchange that carried it, where
it is true. The per-page accumulation and raw-entry offset existed only to feed
the removed publish, and go with it.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RDXCSrnSBV7vZfxycG5Z6E
Signed-off-by: cliffhall <cliff@futurescale.com>
Copilot round 9.
The exclusion scan is a separate, uncached tools/list exchange, so its entries
need not be the ones rendered — the server can answer differently, and a
cache-served aggregate is a different list version. It was overwriting the
aggregate's report wholesale, replacing accurate indices with indices into a
list nobody is looking at. The aggregate now owns the report; the scan
publishes only when the aggregate left nothing, which is the cache-served case
that made it publish at all (round 2).
A conforming re-fetch no longer swallows the mark: the list can change between
the strict page and the lenient one, and a clean second response says nothing
about the first, which was refused. It is marked with the original decode
error when there is nothing per-item to name.
The scan's fallback re-fetch now goes through invokeMcpClient, like the strict
call beside it and like salvageList's walk, so an auth challenge on that
request can still be recovered instead of being swallowed by listAllTools's
best-effort scan catch.
Also fixes listTools dropping an empty-string cursor (truthiness rather than
!== undefined) — its own fallback leg already preserved it, so the two legs of
one scan page disagreed about which page they had fetched.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RDXCSrnSBV7vZfxycG5Z6E
Signed-off-by: cliffhall <cliff@futurescale.com>
Copilot round 8.
The era contract validated with the wire schema but returned the raw input,
so the wire schema's normalization was discarded — both era schemas are
z.object, which strips unknown top-level fields as the strict codec does, so
salvaged tools could carry fields the strict path drops. The parsed output is
now what gets normalized and returned. This also simplifies the bridge: issues
come straight from the wire parse with their paths intact, so the custom-issue
forwarding (and its early describeIssue application) is gone.
The malformed-entry report is written by the aggregate walk's salvage, but in
paginated mode the tools/prompts/resources panels render the paged stores,
which neither write nor clear it — leaving a stale warning above a freshly
fetched page it does not describe, with indices into an aggregate the user is
no longer looking at. Those reports are now suppressed in paginated mode;
resource templates keep theirs, having no paged store. Suppressing rather than
adding per-page salvage: pointing at the wrong entry costs more than staying
quiet, and per-page salvage is a feature, not a display fix.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RDXCSrnSBV7vZfxycG5Z6E
Signed-off-by: cliffhall <cliff@futurescale.com>
Copilot round 7.
The era contract was derived from the neutral ToolSchema plus one override,
which diverged from the real wire schemas both ways. Measured against SDK
2.0.0 on a real connection:
- modern accepts `inputSchema.properties: 5` and `required: [1]`; ours
rejected them, so a malformed sibling made salvage DROP a conforming tool
and report it as malformed — a false accusation from the screen whose job
is to say which entry is wrong.
- legacy rejects `outputSchema.properties: 5` / `required: [1]`; ours
accepted them, readmitting entries the strict path refused.
- modern rejects a non-string `inputSchema.$schema`, legacy accepts it;
ours accepted on both.
Each era is now expressed as its own wire schema, bridged to Tool by
`eraToolItemSchema`. The bridge exists because both eras accept values the
exported Tool type cannot describe (the SDK returns them typed as Tool
anyway), so the wire schema's output type is irrelevant and z.custom<Tool>
supplies the caller's type — which is what avoids an `as unknown as` here.
Forwarded issues keep their path so the report still names the field.
The full measured matrix is pinned table-driven in listSalvage-era.test.ts:
each row asserts the SDK's verdict AND that our contract agrees, so a
divergence that closes or widens fails loudly.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RDXCSrnSBV7vZfxycG5Z6E
Signed-off-by: cliffhall <cliff@futurescale.com>
Copilot round 6.
Bound both fallback walks with LIST_MAX_PAGES, which is now also passed to
the SDK client as `listMaxPages` so the strict aggregate and its fallback
share one value by construction. A server emitting endlessly UNIQUE cursors
never trips the repeated-cursor guard, so the salvage re-walk and the
SEP-2243 exclusion scan would page and accumulate without bound. Hitting the
cap aborts rather than returning a truncated list as a complete one.
Mark the exclusion scan's refused response on every fallback-failure path.
`salvageList` can rethrow bare because ManagedListState marks from the error
it receives; the scan has no such backstop — listAllTools catches and logs a
failing scan by design — so an unmarked rethrow left a response the client
demonstrably refused rendering as a clean exchange.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RDXCSrnSBV7vZfxycG5Z6E
Signed-off-by: cliffhall <cliff@futurescale.com>
Real deployments gate IdP selection, prompt behavior, or audience on a query
parameter the OAuth/OIDC core specs never standardized (Keycloak's
`kc_idp_hint`, OIDC's `login_hint` / `prompt` / `acr_values`, Auth0's
`audience`), and none of them could be driven from the Inspector.
The Inspector never builds the authorize URL — the SDK's `auth()` does, and it
exposes no hook for extra parameters — so the merge happens at the one seam
that sees the finished URL: `BaseOAuthClientProvider.redirectToAuthorization`.
It lives in `core/`, so web, CLI, and TUI all inherit it.
- `core/auth/authorizationParams.ts`: the reserved-key list (`client_id`,
`code_challenge`, `code_challenge_method`, `redirect_uri`, `resource`,
`response_type`, `scope`, `state`), the key validator the form renders
inline, and the merge itself. Reserved keys are dropped with a warning
rather than silently overriding PKCE / the state binding / RFC 8707;
blank-key rows are skipped; the caller's URL is never mutated.
- Applied to the authorization request only — nothing wires it into the token
request.
- Persisted as `oauth.authorizationParams` (a string record) on the per-server
config beside `scopes`, edited as key/value rows in Server Settings ->
Authorization ("Additional authorization parameters"), with an inline error
per offending row plus a section-level warning.
- The enterprise-managed (EMA) leg redirects to a *different* authorization
server (the enterprise IdP), so it goes through the explicit
`redirectToExternalAuthorization` entry point, which skips the merge — a
`kc_idp_hint` or `audience` meant for the MCP server's AS must never be
appended to the IdP's authorize URL.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RDXCSrnSBV7vZfxycG5Z6E
Signed-off-by: cliffhall <cliff@futurescale.com>
Both findings are the same shape: the fallback accepting what the strict path
would not.
- `ModernResultEnvelopeSchema` typed `ttlMs` as any number, where the codec
requires a non-negative integer. `ttlMs: -1` (or `0.5`) plus one bad entry
would salvage, reporting the entry and hiding the envelope violation. Tightened
to `z.int().min(0)`.
- The item schema was the SDK's era-NEUTRAL `ToolSchema`, and the negotiated
era's wire schema is stricter in ways the SDK does not export. Verified
against SDK 2.0.0 on a real connection: on legacy, a tool with
`outputSchema: { type: "array" }` is refused by the strict aggregate but
accepted by the neutral schema; on modern, `execution` is STRIPPED by the
strict path but kept by the neutral one. Either way a salvaged list could show
— and let the user call — a tool the connection's era says isn't valid.
`toolItemSchemaForEra` closes both. Replicating rules that live in the SDK's
internals is a real cost; `listSalvage-era.test.ts` is what makes it safe to
pay, pinning each rule to the strict path's OBSERVED behavior so a divergence
that closes or widens fails loudly instead of rotting.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HDdo1rNRVQnVRRdsqSrvG3
Signed-off-by: cliffhall <cliff@futurescale.com>
Copilot round 4: no new visible comments, two in the suppressed block. Both
valid.
1. The `type: [T, "null"]` encoding keeps its keywords at the top level, so a
nullable enum written that way carries the null *inside* the list:
`{ type: ["string", "null"], enum: ["envio", "recebimento", null] }`. The
collapse left it there, which broke both renderers in different ways — the
web dispatcher handed `null` to Mantine as option data, and the TUI's
all-strings check rejected the whole enum and fell back to a plain text
field. Neither is the dropdown the schema asks for.
The collapse now strips a `null` member whenever it flattens, since that
fact has already moved onto `nullable`. An enum of nothing but `null`
drops out entirely: it offers no value a dropdown could show, so the plain
widget serves the field better than an empty select. Covered in core and
end-to-end in both dispatchers.
2. The new integration test claimed in its own header to pin the registry and
config wiring, then called `createNullableFieldsTool()` directly — so a
misspelt preset case or a config naming a dead preset would have left it
green. A fair "this does not test what it says" catch. It now resolves the
checked-in `nullable-fields-http.json` through `loadConfig` →
`resolveConfig`, matching `duplicate-tool-names.test.ts`, and asserts the
resolved tool names before booting the server.
Signed-off-by: cliffhall <cliff@futurescale.com>
Two findings, both about the fallback accepting more than it should.
- The lenient page schema validated pagination and nothing else, so a response
carrying a malformed entry AND an unrelated top-level violation salvaged: the
entry was reported and the other violation vanished with the strict error that
caught it. The page schema is now DERIVED from the method's real result schema
with only its item array relaxed (`lenientListPageSchema`), so every other
field — `_meta`, `nextCursor`, anything a future revision adds — is still
validated. The modern era envelope the raw-wire path bypasses (`resultType`,
`ttlMs`, `cacheScope`) is checked explicitly for the same reason.
Salvage is also narrowed to `InvalidResult`: `UnsupportedResultType` means the
codec doesn't recognize what KIND of frame arrived, so it isn't a list with a
bad entry and reporting an item-level reason would hide a protocol failure.
Marking the Protocol entry still applies to both, hence two predicates.
- On every rethrow path the re-walk had already answered the method, so the
caller's own `markResponseRejected` (in `ManagedListState.refresh`) landed on
the lenient page that succeeded and left the refused response clean. The
correlation is restored before rethrowing — deleted when nothing had answered
before, so the outer mark is a no-op rather than a lie.
`nextCursorOf` / `rawItemsOf` narrow the page's fields instead of casting: the
runtime-keyed `extend()` loses the shape, and asserting it back would be exactly
the unjustified cast the last round removed.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HDdo1rNRVQnVRRdsqSrvG3
Signed-off-by: cliffhall <cliff@futurescale.com>
Three findings (one inline, two in the suppressed block), all valid.
1. `toConstOptions` still stringified every non-null `const`, so
`items: { anyOf: [{ const: 1 }, { const: 2 }] }` submitted `["1"]` where
the schema says `[1]`. Same wrong-type-on-the-wire problem that keeps a
numeric `enum` off the select path, so it gets the same answer: only a
string `const` is selectable, anything else falls to the JSON editor
where the value keeps its type.
2. `hasMissingRequiredFields` derived null admission from the *renderer's*
collapse, which is narrower on purpose — it only flattens a two-member
union. So `anyOf: [string, number, null]` renders through the JSON
fallback (where `null` is typeable) yet was still marked missing,
disabling submit on a value the schema accepts. Null admission is a
validity question, not a rendering one, so it is now its own predicate:
`admitsNull` in core, recognizing `nullable`, `type: "null"`,
`type: [..., "null"]`, and a null branch anywhere in an `anyOf`/`oneOf`
of any size.
3. Nothing exercised the `record_shipment` preset — the web and TUI tests
build the `anyOf` shape by hand, which verifies the renderers but
*assumes* the premise the whole fix rests on: that Zod's `.nullish()`
emits `anyOf: [<branch>, { type: "null" }]` with the enum on the branch.
A Zod output change or a registry typo would leave every unit test green
while the documented reproduction server quietly stopped reproducing.
Added an integration test that boots the preset over HTTP, asserts the
wire shape, and runs it back through the collapse — following the
`duplicate-tool-names.test.ts` precedent, which exists for the same
reason (#1957).
Signed-off-by: cliffhall <cliff@futurescale.com>
The visible review was clean; all five findings were in the suppressed block,
two of them real defects.
- The exclusion scan's aggregate offset was wrong twice over: `excluded.length`
counts only invalid-header tools, and on page one it also folded in the
current page's valid count. A reported index therefore pointed at a position
that doesn't exist — worst for an unlabeled entry, whose index is all the user
has. Replaced with a raw-entry offset advanced by each page's valid + malformed
counts.
- The scan's own refused response was never marked. It is a distinct exchange
from the aggregate's, and the lenient re-fetch moves the method correlation
off it, so its Protocol entry rendered as a clean success the client had in
fact refused. Marked with the captured id, as `salvageList` already does.
- Three double casts in the test fixture are gone: `ListPage` now models raw
wire data (`items: unknown`, `nextCursor?: unknown`), which is what a server
built to emit spec-forbidden shapes actually serves. Declaring the validated
type and casting malformed values through it asserted the opposite of what
these tests exist to check. The cursor map narrows to `string` at the one
place a cursor addresses a page.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HDdo1rNRVQnVRRdsqSrvG3
Signed-off-by: cliffhall <cliff@futurescale.com>
All four findings taken (one visible, three in the suppressed block), plus the
`excludedTools` sibling leak the round-1 reply had deferred:
- `rawItemsOf` read a missing or non-array list member as an empty page, so a
top-level violation (`{ tools: "invalid" }`) on a later page was discarded
and the call returned a silently truncated list. It now returns `undefined`
for "not a list at all", and the walk rethrows the original strict error.
- A failure of the re-walk ITSELF (a malformed `nextCursor`, a connection lost
mid-walk) escaped in place of the original error, whose response is the one
`rejectedResponseId` was captured for. The walk is now wrapped: the original
surfaces, the secondary is logged so it stays diagnosable.
- The exclusion scan's lenient fallback discarded its malformed entries despite
the comment claiming they were reported. Observable when the aggregate is
served from cache: the salvage never runs, so nothing else reports them. The
scan now returns them and `refreshExcludedTools` reports them — only ever
adding, so a clean scan can't clear what the salvage path just wrote.
- An empty-string cursor is a valid cursor, and `cursor ? {cursor} : {}` dropped
it, re-requesting page one and duplicating its entries until the
repeated-cursor guard fired. Both walks now test `!== undefined`.
- `excludedTools` had the same session leak fixed for `malformedListItems` last
round: cleared only on an explicit `disconnect()`, so the `onerror`-with-no-
`onclose` route carried the previous server's set into the next connection.
Cleared on the start-clean path too — safe because every `listAllTools`
recomputes it, so this can only drop a stale set.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HDdo1rNRVQnVRRdsqSrvG3
Signed-off-by: cliffhall <cliff@futurescale.com>
Three real defects it caught, plus the tests it asked for:
- The strict `tools/list` aggregate is filtered by the SDK for invalid
`x-mcp-header` annotations (SEP-2243), but the salvage path was not: one
ordinary malformed tool would trip the fallback and readmit every
invalid-header tool beside it. Salvaged tools now go through the same scan.
- `markResponseRejected` resolves "the last response received for this method",
and the re-walk answers that same method — so the mark landed on a salvage
page that succeeded while the actually-invalid response rendered clean. The
correlation id is now captured before the re-walk.
- The report was cleared only on an explicit `disconnect()`. The route out that
tears down nothing (`onerror` with no `onclose`) carried it into the next
connection, so it is now cleared on the start-clean `resetSessionState()` path
as well.
Fixing the first exposed two more, both found by the new test:
- The modern (2026-07-28) codec validates a result against the spec schema
before the caller's, so a lenient `client.request` is rejected exactly as the
strict one was — there was no salvage on the modern leg at all. It now goes
through the raw-wire channel that exists for the modern `tasks/*` methods.
`withModernEnvelope` is split out of `withModernTaskEnvelope` so a list
request doesn't claim the tasks extension the user may have disabled.
- A malformed tool also blanked the SEP-2243 excluded set, leaving a tool that
vanished for an invalid header with no explanation — #1909's own complaint one
level down. The scan walk now tolerates a malformed sibling.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HDdo1rNRVQnVRRdsqSrvG3
Signed-off-by: cliffhall <cliff@futurescale.com>
The SDK validates a list result as a whole, so a single non-conforming entry
rejected the entire response and the panel showed "Couldn't load resources"
with zero items — the exact opposite of what a debugging tool should do with a
server that is 90% right. Reported against a PHP server whose empty
`annotations` object reaches the wire as `[]` rather than `{}`.
`listAll*` now falls back to a lenient re-walk when — and only when — the SDK
rejects a result it received (`InvalidResult` / `UnsupportedResultType`), keeps
the entries that validate, and reports the rest. A conforming server takes the
strict, cache-aware path exactly as before.
Salvaging is not accepting: the dropped entries are named with their reason in
a warning above the list, and the Protocol entry is still marked rejected, so
the wire truth survives. When every entry validates, the original error is
rethrown rather than swallowed — the rejection was about something the
per-item pass cannot explain.
`isClientDecodeRejection` moves to the new module and is now shared with
managedListState, which had the only copy.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HDdo1rNRVQnVRRdsqSrvG3
Signed-off-by: cliffhall <cliff@futurescale.com>
- The `retriedDynamic` latch test never made the retry fail: a real `listen(0)`
always succeeds, so it produced one EADDRINUSE and passed with the latch
deleted. Mock `node:http` so every listen fails, asserting exactly two listen
attempts and a settled result. Verified by mutation: removing the latch now
hangs the test instead of passing it.
- The migration guide's Docker notes still described a dynamic sandbox port and
told readers to pin a custom one, contradicting the new fixed 6275 two
sections later.
- Narrowed "or any container" in the troubleshooting entry, which overstated the
scope and contradicted the PR's own scope note. It now names the IPv4-forwarder
case and explicitly excludes the official image and a plain loopback -p.
Signed-off-by: cliffhall <cliff@futurescale.com>
Two defaults that each break the Inspector inside a container.
#1951 — the web UI never loads in a VS Code dev container. `HOST` defaulted
to the NAME `localhost`, and `server.listen(port, host)` resolves a name
through `dns.lookup` and binds the SINGLE address it gets back. Since Node 17
the default result order is `verbatim`, so on a glibc Linux host — whose
/etc/hosts maps localhost to both 127.0.0.1 and ::1 — that is `::1`. The
server bound IPv6 loopback only and refused every IPv4 client, including the
dev container port forwarder. It presents as an endless spinner rather than a
connection error because the forwarder accepts on the host side before failing
to connect inward. Default to the ADDRESS 127.0.0.1 instead; an explicit
HOST=localhost is still honored as typed.
#2008 — the Apps tab is unreachable in any container. The MCP Apps sandbox
listened on an OS-assigned port, so it changed every run and could never be
named in a devcontainer `forwardPorts`, a `docker run -p`, or a tunnel. Default
it to a fixed 6275, mirroring CLIENT_PORT's fixed 6274, and retry once on an
OS-assigned port (loudly) if it is taken — the collision a dynamic default
never had.
Verified in the reporter's exact image
(mcr.microsoft.com/devcontainers/typescript-node:4-24-trixie): both listeners
come up on 127.0.0.1 and answer 200 on 6274 and 6275/sandbox.
Also EXPOSE 6275 and set MCP_SANDBOX_PORT in the image, and document both
defaults in the web README, the migration guide, and the root README's Docker
section.
Signed-off-by: cliffhall <cliff@futurescale.com>
Review feedback on #1987: `waitForDeadTransport` polled `/api/mcp/send`,
which answers `transport_error` from two branches — the `isTransportDead()`
short-circuit and the `catch` around a rejected `send()`. Only the first
implies the transport is marked dead, so the stop condition could not tell
them apart from the response alone.
That was safe in practice: `StdioClientTransport.send` rejects only when
`_process` is undefined, and the transport clears `_process` and calls
`onclose` in one synchronous block, so the `catch` branch is not observable
before `markTransportDead()` has run (confirmed empirically — 0/25 probes
returned early). But the safety rested on an SDK-internal ordering this test
neither states nor controls, and if it changed the send test would silently
cover the `catch` branch instead of the short-circuit it names — a regression
with no failing run attached to it.
Poll `/api/mcp/auth-state` instead. For a session connected without an
authState there is no auth provider, so `setAuthState` always throws and
`transport_error` there has exactly one source: the dead-transport
short-circuit. The route never awaits a transport response, so it cannot
hang either. Same cost, exact condition, no dependence on SDK internals.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WLYmVPVLdeYpE6aMTUsCks
