* feat(servers): move oauthClientSecret and stdio env values to OS keychain (#1356) mcp.json is designed to be tool-shareable (symlinked from Claude Desktop's config, pasted into bug reports, synced via dotfiles), so storing OAuth client secrets and stdio env values in plaintext there meant any of those flows could leak them. Lift both into the OS keychain via @napi-rs/keyring (active replacement for the archived keytar). The wire shape is unchanged: the GET /api/servers handler rehydrates from the keychain so browser code sees the same JSON; the on-disk file no longer contains the secret material. Includes an idempotent migration that lifts plaintext from older mcp.json files (or hand-edited ones) into the keychain on first read. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * fix(servers): tolerate keychain unavailability on non-set ops (#1356) CI on Linux failed because `@napi-rs/keyring` hard-errors without libsecret, and the route handlers were touching the keychain even for no-secret flows (defensive `deleteAllForServer` on POST/DELETE, sweep on PUT). The user's intent for the Linux fallback was "hard fail when a secret is actually involved" — not on every keychain touch. Make `KeyringSecretStore.get / delete / deleteAllForServer` silently degrade when the keychain is unavailable (return null / no-op); `set` remains the one operation that hard-fails with `KeychainUnavailableError`. The migration path catches the error and preserves the disk plaintext rather than losing the secret. Also inject `InMemorySecretStore` in `useServers.test.tsx` and `servers-events.test.ts` so tests never touch a developer's real keychain even when libsecret is present. Adds `KeyringSecretStore` tests (vi.mock'd native bindings) and an integration suite that simulates Linux-without-libsecret to verify the tolerance contract end-to-end. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * fix(servers): close GET migration race; reverse keychain/disk order (#1356) Addresses code review: 1. **Concurrency**: GET /api/servers was reading + migrating outside the write lock, then writing inside it — a concurrent POST/PUT/DELETE between the read and the lock acquisition could be clobbered by the migration write. Also affected the first-launch seed write. Fix: take a fast unlocked peek; if the file is absent (seed branch) or carries plaintext to migrate (migration branch), re-read inside the write lock and decide there. Added `hasPlaintextSecrets()` as a pure predicate so the common no-migration path stays lock-free. 2. **Write ordering**: POST and PUT wrote disk first, then keychain. A keychain `set` failure mid-write left a half-configured entry on disk and trapped retries at 409. Reversed to keychain-first so a 503 leaves both stores in their pre-write state. PUT in-place now sets new fields, writes disk, then deletes obsolete fields (a failed disk write leaves orphan keychain entries — recoverable — rather than premature deletions of fields the user still expects to see). 3. **Tolerance simplification**: dropped the no-entry regex in `KeyringSecretStore.delete` — both "no entry" and "keychain unavailable" collapse to the same desired outcome (the entry isn't there anymore), so a uniform swallow is clearer. 4. **Parallelism**: `readKeychainEntriesFor` now `Promise.all`s the per-field `get` calls. macOS Keychain round-trips are 10-50ms; serializing 5+ env vars per server × N servers is a meaningful wall-clock cost on GET. 5. **Spec doc**: reflects the tolerance contract (only `set` hard-fails) and the keychain-before-disk write ordering rationale. New regression test in servers-route.test.ts covers the retry-after-503 contract directly. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * chore(servers): parallel keychain writes + migration partial-state comment (#1356) Two nice-to-have observations from the re-review: - `writeKeychainEntriesFor` now `Promise.all`s its per-field sets, matching the read side. No ordering requirement among distinct (id, field) keys; on unavailability every set rejects and Promise.all surfaces the first rejection (still 503 from the route handler). - `migratePlaintextSecrets` catch block now spells out the partial-migration semantics: we return the original config so the disk plaintext stays intact, and the next successful GET retries with the keychain-wins branch absorbing already-set entries. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * chore(servers): parallel keychain deletes for symmetry (#1356) Picks up the symmetry observation from the third review: `deleteKeychainFields` now uses `Promise.all` like its read and write siblings. Distinct (id, field) deletes have no ordering requirement, and `secretStore.delete` is already a silent no-op on unavailability so Promise.all has no failure-mode surprise. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
Cliff Hall
GitHub
parent
a9df501856
commit
037225a9fb
4 changed files with 2013 additions and 24 deletions
179
core/auth/node/secret-store.ts
Normal file
179
core/auth/node/secret-store.ts
Normal file
|
|
@ -0,0 +1,179 @@
|
|||
/**
|
||||
* Per-server secret storage backed by the OS keychain.
|
||||
*
|
||||
* Service name `mcp-inspector`; account `${serverId}:${field}`. Fields used
|
||||
* by the current code: `oauth-client-secret` and `env:${KEY}` (one per
|
||||
* stdio env variable). Keeping the account namespaced by `serverId` lets
|
||||
* us drop every entry for a server in one sweep when DELETE
|
||||
* /api/servers/:id runs, and lets `findCredentials(SERVICE)` enumerate
|
||||
* everything we own for migration / debugging.
|
||||
*
|
||||
* Node-only — `@napi-rs/keyring` uses native bindings (Keychain Services
|
||||
* on macOS, Credential Manager on Windows, libsecret on Linux). The
|
||||
* browser side never imports this; it gets values rehydrated into the
|
||||
* `/api/servers` response by the Hono handler.
|
||||
*/
|
||||
import {
|
||||
AsyncEntry,
|
||||
findCredentialsAsync,
|
||||
} from "@napi-rs/keyring";
|
||||
|
||||
const SERVICE_NAME = "mcp-inspector";
|
||||
|
||||
export {
|
||||
SECRET_FIELD_OAUTH_CLIENT_SECRET,
|
||||
envSecretField,
|
||||
} from "../secret-fields.js";
|
||||
|
||||
/** Parse a stored account key back into its server id and field. */
|
||||
export function parseAccount(
|
||||
account: string,
|
||||
): { serverId: string; field: string } | null {
|
||||
const idx = account.indexOf(":");
|
||||
if (idx <= 0 || idx === account.length - 1) return null;
|
||||
return {
|
||||
serverId: account.slice(0, idx),
|
||||
field: account.slice(idx + 1),
|
||||
};
|
||||
}
|
||||
|
||||
const buildAccount = (serverId: string, field: string): string =>
|
||||
`${serverId}:${field}`;
|
||||
|
||||
/**
|
||||
* Thrown when the OS keychain is unavailable — typically Linux without
|
||||
* libsecret / gnome-keyring installed. Surfaced as a 503 by the API
|
||||
* handlers so the UI can show an actionable error rather than a generic
|
||||
* 500. macOS and Windows always have a working keychain, so this only
|
||||
* realistically fires on minimal Linux installs.
|
||||
*/
|
||||
export class KeychainUnavailableError extends Error {
|
||||
constructor(cause: unknown) {
|
||||
super(
|
||||
`OS keychain is not available. On Linux, install libsecret / gnome-keyring. ` +
|
||||
`Underlying error: ${cause instanceof Error ? cause.message : String(cause)}`,
|
||||
);
|
||||
this.name = "KeychainUnavailableError";
|
||||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* Storage interface for the per-server secrets we lift off
|
||||
* `~/.mcp-inspector/mcp.json`. Implemented by `KeyringSecretStore` (the
|
||||
* production impl) and `InMemorySecretStore` (used in tests so the suite
|
||||
* doesn't require libsecret in CI).
|
||||
*/
|
||||
export interface SecretStore {
|
||||
get(serverId: string, field: string): Promise<string | null>;
|
||||
set(serverId: string, field: string, value: string): Promise<void>;
|
||||
/** No-op if no entry exists. */
|
||||
delete(serverId: string, field: string): Promise<void>;
|
||||
/** Remove every secret stored for this server id (called on DELETE /api/servers/:id). */
|
||||
deleteAllForServer(serverId: string): Promise<void>;
|
||||
}
|
||||
|
||||
/**
|
||||
* Default implementation. Each operation constructs a fresh `AsyncEntry`;
|
||||
* the native side is cheap and the alternative (caching entries by
|
||||
* (serverId, field)) just trades native-handle bookkeeping for an
|
||||
* allocation that's measured in microseconds. `getPassword` returns
|
||||
* `undefined` for a missing entry — we normalize to `null` so callers
|
||||
* can use `=== null` rather than truthiness (an empty-string secret is
|
||||
* a real value and must round-trip).
|
||||
*
|
||||
* **Availability behavior.** When the keychain is unavailable (the
|
||||
* typical case is Linux without libsecret / gnome-keyring), `set` is
|
||||
* the only operation that throws `KeychainUnavailableError` — that's
|
||||
* the moment where data would actually be lost. `get` returns `null`
|
||||
* (as if no entry existed) and the destructive operations silently
|
||||
* no-op (there's nothing to delete anyway). This keeps non-secret
|
||||
* flows working on a stock CI runner / minimal Linux box; the user
|
||||
* only hits a hard error when they actually try to save a secret.
|
||||
*/
|
||||
export class KeyringSecretStore implements SecretStore {
|
||||
async get(serverId: string, field: string): Promise<string | null> {
|
||||
const entry = new AsyncEntry(SERVICE_NAME, buildAccount(serverId, field));
|
||||
try {
|
||||
const v = await entry.getPassword();
|
||||
return v ?? null;
|
||||
} catch {
|
||||
// Tolerate keychain unavailability on reads: there's no value to
|
||||
// surface either way. Hard-failing here would break GET flows
|
||||
// that don't touch any secret material (most of the test suite,
|
||||
// and most user sessions on a Linux box without libsecret).
|
||||
return null;
|
||||
}
|
||||
}
|
||||
|
||||
async set(serverId: string, field: string, value: string): Promise<void> {
|
||||
const entry = new AsyncEntry(SERVICE_NAME, buildAccount(serverId, field));
|
||||
try {
|
||||
await entry.setPassword(value);
|
||||
} catch (err) {
|
||||
// The only operation that hard-fails — if we can't persist the
|
||||
// secret, the user needs to know now rather than discover later
|
||||
// that their value disappeared. Routes translate this to a 503.
|
||||
throw new KeychainUnavailableError(err);
|
||||
}
|
||||
}
|
||||
|
||||
async delete(serverId: string, field: string): Promise<void> {
|
||||
const entry = new AsyncEntry(SERVICE_NAME, buildAccount(serverId, field));
|
||||
try {
|
||||
await entry.deleteCredential();
|
||||
} catch {
|
||||
// Both reasons for a throw collapse to the same desired outcome
|
||||
// ("the entry isn't there anymore"): `deleteCredential` raises
|
||||
// NoEntry for a missing credential, and the native binding
|
||||
// raises a runtime error when the keychain itself is unavailable.
|
||||
// We treat both as success — there's no value to lose either
|
||||
// way, and `set` is the operation that hard-fails when the
|
||||
// keychain is actually down.
|
||||
}
|
||||
}
|
||||
|
||||
async deleteAllForServer(serverId: string): Promise<void> {
|
||||
let creds: Array<{ account: string; password: string }>;
|
||||
try {
|
||||
creds = await findCredentialsAsync(SERVICE_NAME);
|
||||
} catch {
|
||||
// Same reasoning as `delete`: nothing was written, nothing to sweep.
|
||||
return;
|
||||
}
|
||||
const prefix = `${serverId}:`;
|
||||
for (const c of creds) {
|
||||
if (!c.account.startsWith(prefix)) continue;
|
||||
const parsed = parseAccount(c.account);
|
||||
if (!parsed || parsed.serverId !== serverId) continue;
|
||||
await this.delete(serverId, parsed.field);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* Test double — substituted via the `secretStore` option on the remote
|
||||
* server factory. Mirrors the keyring contract exactly so swapping it
|
||||
* in/out doesn't change behavior beyond persistence.
|
||||
*/
|
||||
export class InMemorySecretStore implements SecretStore {
|
||||
private readonly map = new Map<string, string>();
|
||||
|
||||
async get(serverId: string, field: string): Promise<string | null> {
|
||||
return this.map.get(buildAccount(serverId, field)) ?? null;
|
||||
}
|
||||
|
||||
async set(serverId: string, field: string, value: string): Promise<void> {
|
||||
this.map.set(buildAccount(serverId, field), value);
|
||||
}
|
||||
|
||||
async delete(serverId: string, field: string): Promise<void> {
|
||||
this.map.delete(buildAccount(serverId, field));
|
||||
}
|
||||
|
||||
async deleteAllForServer(serverId: string): Promise<void> {
|
||||
const prefix = `${serverId}:`;
|
||||
for (const key of [...this.map.keys()]) {
|
||||
if (key.startsWith(prefix)) this.map.delete(key);
|
||||
}
|
||||
}
|
||||
}
|
||||
14
core/auth/secret-fields.ts
Normal file
14
core/auth/secret-fields.ts
Normal file
|
|
@ -0,0 +1,14 @@
|
|||
/**
|
||||
* Field identifiers for the per-server values held in the OS keychain.
|
||||
*
|
||||
* Pure constants — kept out of `core/auth/node/` so the converter in
|
||||
* `core/mcp/serverList.ts` (browser-safe) and the Node-only keychain
|
||||
* backend in `core/auth/node/secret-store.ts` agree on the spelling
|
||||
* without dragging the native binding import into browser code.
|
||||
*/
|
||||
|
||||
/** Field name for an entry's OAuth client secret. */
|
||||
export const SECRET_FIELD_OAUTH_CLIENT_SECRET = "oauth-client-secret";
|
||||
|
||||
/** Field name for one stdio env-variable value. */
|
||||
export const envSecretField = (envKey: string): string => `env:${envKey}`;
|
||||
|
|
@ -33,8 +33,11 @@ import type {
|
|||
} from "../../types.js";
|
||||
import {
|
||||
DEFAULT_SEED_CONFIG,
|
||||
expectedSecretFields,
|
||||
extractSecretsFromStored,
|
||||
INSPECTOR_FIELD_KEYS,
|
||||
inspectorSettingsToStoredFields,
|
||||
mergeSecretsIntoStored,
|
||||
normalizeServerType,
|
||||
storedFieldsToInspectorSettings,
|
||||
stripInspectorFields,
|
||||
|
|
@ -42,6 +45,11 @@ import {
|
|||
import { RemoteSession } from "./remote-session.js";
|
||||
import { createTokenAuthProvider } from "./tokenAuthProvider.js";
|
||||
import { API_SERVER_ENV_VARS } from "../constants.js";
|
||||
import {
|
||||
KeychainUnavailableError,
|
||||
KeyringSecretStore,
|
||||
type SecretStore,
|
||||
} from "../../../auth/node/secret-store.js";
|
||||
|
||||
/**
|
||||
* Shape of the initial config returned by GET /api/config (defaults for client).
|
||||
|
|
@ -83,6 +91,15 @@ export interface RemoteServerOptions {
|
|||
|
||||
/** Initial config for GET /api/config. Caller must pass this (e.g. from webServerConfigToInitialPayload(config)). */
|
||||
initialConfig: InitialConfigPayload;
|
||||
|
||||
/**
|
||||
* Backend for per-server secret values that we keep out of mcp.json
|
||||
* (OAuth client secret, stdio env values). Defaults to a
|
||||
* `KeyringSecretStore` that talks to the OS keychain via
|
||||
* `@napi-rs/keyring`. Tests inject `InMemorySecretStore` so the suite
|
||||
* doesn't need libsecret on Linux CI runners.
|
||||
*/
|
||||
secretStore?: SecretStore;
|
||||
}
|
||||
|
||||
export interface CreateRemoteAppResult {
|
||||
|
|
@ -277,6 +294,8 @@ export function createRemoteApp(
|
|||
const { logger: fileLogger, allowedOrigins } = options;
|
||||
const storageDir = options.storageDir ?? getDefaultStorageDir();
|
||||
const mcpConfigPath = options.mcpConfigPath ?? getDefaultMcpConfigPath();
|
||||
const secretStore: SecretStore =
|
||||
options.secretStore ?? new KeyringSecretStore();
|
||||
|
||||
// --- /api/servers/events: file-watch fanout state -----------------------
|
||||
//
|
||||
|
|
@ -1120,16 +1139,234 @@ export function createRemoteApp(
|
|||
return { mcpServers: normalizeMcpServers(parsed?.mcpServers) };
|
||||
};
|
||||
|
||||
// ---- Keychain helpers -------------------------------------------------
|
||||
//
|
||||
// Centralizes the "write the secrets in this entry to the keychain" and
|
||||
// "fetch all secrets for this entry from the keychain" steps so the
|
||||
// POST/PUT/DELETE/GET handlers stay readable. Each operation translates a
|
||||
// `KeychainUnavailableError` from the underlying store into a Hono 503
|
||||
// — the only realistic trigger is Linux without libsecret, and the user
|
||||
// can install it without restarting.
|
||||
|
||||
// Same Promise.all reasoning as `readKeychainEntriesFor`: each set
|
||||
// is a native round-trip and there's no ordering requirement among
|
||||
// distinct (id, field) keys. If the keychain is unavailable every
|
||||
// set will reject — Promise.all surfaces the first rejection, which
|
||||
// is the right signal for the route handler (translates to 503).
|
||||
const writeKeychainEntriesFor = async (
|
||||
id: string,
|
||||
secrets: Record<string, string>,
|
||||
): Promise<void> => {
|
||||
await Promise.all(
|
||||
Object.entries(secrets).map(([field, value]) =>
|
||||
secretStore.set(id, field, value),
|
||||
),
|
||||
);
|
||||
};
|
||||
|
||||
// Fetch every keychain value an entry could need into a flat record
|
||||
// keyed by field name. Used by the GET handler to rehydrate the on-disk
|
||||
// stripped shape into the shape today's browser code expects. Missing
|
||||
// fields are simply absent from the result (so `mergeSecretsIntoStored`
|
||||
// leaves the disk value untouched).
|
||||
// Issue many keychain reads in parallel. On macOS each round-trip to
|
||||
// Keychain Services is 10-50ms; a server with 20 entries × 5 env vars
|
||||
// each would otherwise serialize to ~5s of rehydration on every GET.
|
||||
// `@napi-rs/keyring`'s async APIs release the event loop, so
|
||||
// Promise.all is a real win, not just stylistic.
|
||||
const readKeychainEntriesFor = async (
|
||||
id: string,
|
||||
fields: string[],
|
||||
): Promise<Record<string, string>> => {
|
||||
const values = await Promise.all(
|
||||
fields.map(async (field) => [field, await secretStore.get(id, field)] as const),
|
||||
);
|
||||
const out: Record<string, string> = {};
|
||||
for (const [field, v] of values) {
|
||||
if (v !== null) out[field] = v;
|
||||
}
|
||||
return out;
|
||||
};
|
||||
|
||||
// Cheap structural check used by the GET handler to decide whether
|
||||
// to enter the write lock for migration. Pure — no keychain or disk
|
||||
// access. The actual migration (`migratePlaintextSecrets`) writes
|
||||
// to the keychain, so we must not run it speculatively outside the
|
||||
// lock; this predicate lets us keep the fast path lock-free.
|
||||
const hasPlaintextSecrets = (config: MCPConfig): boolean => {
|
||||
for (const stored of Object.values(config.mcpServers)) {
|
||||
const { secrets } = extractSecretsFromStored(stored);
|
||||
if (Object.keys(secrets).length > 0) return true;
|
||||
}
|
||||
return false;
|
||||
};
|
||||
|
||||
// Migrate plaintext secrets in a freshly-read mcp.json into the
|
||||
// keychain. Idempotent: when the keychain already has a value for
|
||||
// `(id, field)`, that value wins and the disk plaintext is dropped
|
||||
// unread. Returns the rewritten config plus a flag so the GET handler
|
||||
// knows whether to persist the cleanup.
|
||||
//
|
||||
// When the keychain is unavailable (Linux without libsecret), the
|
||||
// first `secretStore.set` throws `KeychainUnavailableError`. We catch
|
||||
// it and abandon the migration for this GET — the on-disk file stays
|
||||
// as-is so the user's secret isn't lost, and the next GET retries
|
||||
// (e.g. after the user installs libsecret).
|
||||
const migratePlaintextSecrets = async (
|
||||
config: MCPConfig,
|
||||
): Promise<{ migrated: MCPConfig; changed: boolean }> => {
|
||||
let changed = false;
|
||||
const next: MCPConfig = { mcpServers: {} };
|
||||
try {
|
||||
for (const [id, stored] of Object.entries(config.mcpServers)) {
|
||||
const { stripped, secrets } = extractSecretsFromStored(stored);
|
||||
if (Object.keys(secrets).length === 0) {
|
||||
next.mcpServers[id] = stored;
|
||||
continue;
|
||||
}
|
||||
for (const [field, value] of Object.entries(secrets)) {
|
||||
const existing = await secretStore.get(id, field);
|
||||
if (existing === null) {
|
||||
await secretStore.set(id, field, value);
|
||||
}
|
||||
// existing !== null → keychain already had a value for this
|
||||
// (id, field); keep the keychain authoritative and drop the
|
||||
// plaintext from disk. This handles the case where a user has
|
||||
// edited mcp.json by hand after the original migration.
|
||||
}
|
||||
next.mcpServers[id] = stripped;
|
||||
changed = true;
|
||||
}
|
||||
} catch (err) {
|
||||
if (err instanceof KeychainUnavailableError) {
|
||||
if (fileLogger) {
|
||||
fileLogger.warn(
|
||||
{ err: err.message },
|
||||
"Keychain unavailable; skipping plaintext-secret migration on this read. Existing mcp.json plaintext values are preserved.",
|
||||
);
|
||||
}
|
||||
// Partial-migration semantics: if `set` threw partway through
|
||||
// the loop, some servers may already have keychain entries
|
||||
// while others don't. We deliberately return the original
|
||||
// `config` (not the partial `next`) so the disk file stays
|
||||
// intact — the next successful GET runs the migration again,
|
||||
// and the idempotent "keychain wins on conflict" branch
|
||||
// (`existing !== null`) silently absorbs the already-set
|
||||
// entries. No data loss; one wasted set on retry per
|
||||
// already-migrated entry.
|
||||
return { migrated: config, changed: false };
|
||||
}
|
||||
throw err;
|
||||
}
|
||||
return { migrated: next, changed };
|
||||
};
|
||||
|
||||
// Rehydrate every server's secrets so the GET response matches what
|
||||
// the browser saw before the keychain split. Runs after migration in
|
||||
// the GET handler.
|
||||
const rehydrateConfig = async (config: MCPConfig): Promise<MCPConfig> => {
|
||||
const out: MCPConfig = { mcpServers: {} };
|
||||
for (const [id, stored] of Object.entries(config.mcpServers)) {
|
||||
const fields = expectedSecretFields(stored);
|
||||
const secrets = await readKeychainEntriesFor(id, fields);
|
||||
out.mcpServers[id] = mergeSecretsIntoStored(stored, secrets);
|
||||
}
|
||||
return out;
|
||||
};
|
||||
|
||||
// Fields the previous entry held that the new entry doesn't —
|
||||
// candidates for deletion after the disk write succeeds. Returned
|
||||
// as an array so the caller can decide where in the write sequence
|
||||
// to perform the destructive operation (we want it after the disk
|
||||
// write so a failed disk write doesn't leave the user with their
|
||||
// old config on disk but missing keychain entries).
|
||||
const computeObsoleteFields = (
|
||||
previousFields: Set<string>,
|
||||
nextSecrets: Record<string, string>,
|
||||
): string[] => {
|
||||
const nextFieldSet = new Set(Object.keys(nextSecrets));
|
||||
const obsolete: string[] = [];
|
||||
for (const field of previousFields) {
|
||||
if (!nextFieldSet.has(field)) obsolete.push(field);
|
||||
}
|
||||
return obsolete;
|
||||
};
|
||||
|
||||
// Parallel for symmetry with `readKeychainEntriesFor` /
|
||||
// `writeKeychainEntriesFor`: distinct (id, field) deletes have no
|
||||
// ordering requirement, and `secretStore.delete` is already a silent
|
||||
// no-op on unavailability so Promise.all has no failure-mode surprise.
|
||||
const deleteKeychainFields = async (
|
||||
id: string,
|
||||
fields: string[],
|
||||
): Promise<void> => {
|
||||
await Promise.all(fields.map((field) => secretStore.delete(id, field)));
|
||||
};
|
||||
|
||||
const keychainErrorResponse = (
|
||||
c: Context,
|
||||
err: unknown,
|
||||
): Response | undefined => {
|
||||
if (err instanceof KeychainUnavailableError) {
|
||||
return c.json({ error: err.message }, 503);
|
||||
}
|
||||
return undefined;
|
||||
};
|
||||
|
||||
app.get("/api/servers", async (c) => {
|
||||
try {
|
||||
// Fast path: peek at the file without taking the write lock. Most
|
||||
// GETs land here — file exists, no plaintext to migrate. The
|
||||
// unlocked read is safe because we don't write anything in this
|
||||
// branch; a concurrent mutation just means the response is a
|
||||
// snapshot from a moment ago.
|
||||
const raw = await readStoreFile(mcpConfigPath);
|
||||
if (raw === null) {
|
||||
await writeMcpAndTrackMtime(serializeStore(DEFAULT_SEED_CONFIG));
|
||||
return c.json(DEFAULT_SEED_CONFIG);
|
||||
if (raw !== null) {
|
||||
const parsed = parseStore(raw) as { mcpServers?: unknown } | null;
|
||||
const onDisk: MCPConfig = {
|
||||
mcpServers: normalizeMcpServers(parsed?.mcpServers),
|
||||
};
|
||||
if (!hasPlaintextSecrets(onDisk)) {
|
||||
return c.json(await rehydrateConfig(onDisk));
|
||||
}
|
||||
}
|
||||
const parsed = parseStore(raw) as { mcpServers?: unknown } | null;
|
||||
return c.json({ mcpServers: normalizeMcpServers(parsed?.mcpServers) });
|
||||
|
||||
// Slow path: either the file is missing (first-launch seed) or it
|
||||
// contains plaintext we need to migrate. Both branches write the
|
||||
// file, so the entire read + decide + write sequence must happen
|
||||
// inside the write lock — otherwise a concurrent POST/PUT/DELETE
|
||||
// could land between our unlocked read and our locked write, and
|
||||
// we'd clobber it. Re-read once we hold the lock so the decision
|
||||
// is based on the same snapshot we're about to mutate.
|
||||
const settled = await withWriteLock(async () => {
|
||||
const rawInside = await readStoreFile(mcpConfigPath);
|
||||
if (rawInside === null) {
|
||||
// Still absent after taking the lock → no concurrent POST
|
||||
// beat us to it; seed the file ourselves. The seed has no
|
||||
// secrets so nothing else to do.
|
||||
await writeMcpAndTrackMtime(serializeStore(DEFAULT_SEED_CONFIG));
|
||||
return DEFAULT_SEED_CONFIG;
|
||||
}
|
||||
const parsedInside = parseStore(rawInside) as
|
||||
| { mcpServers?: unknown }
|
||||
| null;
|
||||
const inside: MCPConfig = {
|
||||
mcpServers: normalizeMcpServers(parsedInside?.mcpServers),
|
||||
};
|
||||
// A concurrent POST may have run between the unlocked peek and
|
||||
// the lock acquisition — if the file now has no plaintext to
|
||||
// migrate, leave it alone and return the latest snapshot.
|
||||
if (!hasPlaintextSecrets(inside)) return inside;
|
||||
const { migrated, changed } = await migratePlaintextSecrets(inside);
|
||||
if (changed) {
|
||||
await writeMcpAndTrackMtime(serializeStore(migrated));
|
||||
}
|
||||
return migrated;
|
||||
});
|
||||
return c.json(await rehydrateConfig(settled));
|
||||
} catch (error) {
|
||||
const keychainResp = keychainErrorResponse(c, error);
|
||||
if (keychainResp) return keychainResp;
|
||||
const msg = error instanceof Error ? error.message : String(error);
|
||||
return c.json({ error: `Failed to read server list: ${msg}` }, 500);
|
||||
}
|
||||
|
|
@ -1174,15 +1411,26 @@ export function createRemoteApp(
|
|||
if (id in current.mcpServers) {
|
||||
return c.json({ error: `Server '${id}' already exists` }, 409);
|
||||
}
|
||||
current.mcpServers[id] = buildStoredEntry(
|
||||
id,
|
||||
body.config,
|
||||
postSettings,
|
||||
);
|
||||
const built = buildStoredEntry(id, body.config, postSettings);
|
||||
// Split secret values out of the new entry — the stripped shape
|
||||
// goes to disk, the values go to the keychain.
|
||||
const { stripped, secrets } = extractSecretsFromStored(built);
|
||||
// Order: sweep → keychain → disk. The keychain write is the
|
||||
// only step that can hard-fail (KeychainUnavailableError on
|
||||
// `set`); doing it before the disk write means a 503 leaves no
|
||||
// disk entry behind, so a retry POST isn't trapped at 409. The
|
||||
// initial sweep handles the case where a previous DELETE failed
|
||||
// midway and left orphans under the same id the user is now
|
||||
// reusing; it's a silent no-op when the keychain is unavailable.
|
||||
await secretStore.deleteAllForServer(id);
|
||||
await writeKeychainEntriesFor(id, secrets);
|
||||
current.mcpServers[id] = stripped;
|
||||
await writeMcpAndTrackMtime(serializeStore(current));
|
||||
return c.json({ ok: true });
|
||||
});
|
||||
} catch (error) {
|
||||
const keychainResp = keychainErrorResponse(c, error);
|
||||
if (keychainResp) return keychainResp;
|
||||
const msg = error instanceof Error ? error.message : String(error);
|
||||
return c.json({ error: `Failed to add server: ${msg}` }, 500);
|
||||
}
|
||||
|
|
@ -1294,22 +1542,56 @@ export function createRemoteApp(
|
|||
nextSettings = settingsIntent.value;
|
||||
break;
|
||||
}
|
||||
// Build the new entry, then split off secrets before writing to
|
||||
// disk. Reconcile uses the previously-stored entry to know which
|
||||
// keychain fields existed so any field the user removed (env
|
||||
// key dropped, OAuth secret cleared) gets cleaned up rather
|
||||
// than orphaned.
|
||||
const built = buildStoredEntry(newId, nextConfig, nextSettings);
|
||||
const { stripped, secrets } = extractSecretsFromStored(built);
|
||||
const next: MCPConfig = { mcpServers: {} };
|
||||
for (const [key, val] of Object.entries(current.mcpServers)) {
|
||||
if (key === originalId) {
|
||||
next.mcpServers[newId] = buildStoredEntry(
|
||||
newId,
|
||||
nextConfig,
|
||||
nextSettings,
|
||||
);
|
||||
next.mcpServers[newId] = stripped;
|
||||
} else {
|
||||
next.mcpServers[key] = val;
|
||||
}
|
||||
}
|
||||
await writeMcpAndTrackMtime(serializeStore(next));
|
||||
// Ordering: write the new keychain entries first, then the
|
||||
// disk file, then clean up obsolete keychain entries.
|
||||
//
|
||||
// - Keychain set is the only hard-fail step (it raises 503 on
|
||||
// `KeychainUnavailableError`). Doing it first means a failed
|
||||
// set leaves both disk and keychain in their pre-PUT state;
|
||||
// the user retries and nothing is half-applied.
|
||||
// - The disk write happens after the keychain is fully primed,
|
||||
// so a successful disk write is also a fully-consistent end
|
||||
// state.
|
||||
// - Obsolete deletion comes last because it's destructive: if
|
||||
// we deleted first and then the disk write failed, the user
|
||||
// would still see the old config on disk but with missing
|
||||
// keychain values. With the current order a failed disk
|
||||
// write leaves orphan keychain entries — recoverable on the
|
||||
// next reconcile or `deleteAllForServer` sweep.
|
||||
if (newId !== originalId) {
|
||||
await writeKeychainEntriesFor(newId, secrets);
|
||||
await writeMcpAndTrackMtime(serializeStore(next));
|
||||
await secretStore.deleteAllForServer(originalId);
|
||||
} else {
|
||||
// In-place update: same id, possibly different fields. Set
|
||||
// the new values first, then write disk, then drop obsolete
|
||||
// fields (env keys the user removed, OAuth secret cleared).
|
||||
const previousFields = new Set(expectedSecretFields(existing));
|
||||
const obsolete = computeObsoleteFields(previousFields, secrets);
|
||||
await writeKeychainEntriesFor(newId, secrets);
|
||||
await writeMcpAndTrackMtime(serializeStore(next));
|
||||
await deleteKeychainFields(newId, obsolete);
|
||||
}
|
||||
return c.json({ ok: true });
|
||||
});
|
||||
} catch (error) {
|
||||
const keychainResp = keychainErrorResponse(c, error);
|
||||
if (keychainResp) return keychainResp;
|
||||
const msg = error instanceof Error ? error.message : String(error);
|
||||
return c.json({ error: `Failed to update server: ${msg}` }, 500);
|
||||
}
|
||||
|
|
@ -1325,13 +1607,20 @@ export function createRemoteApp(
|
|||
return await withWriteLock(async () => {
|
||||
const current = await readMcpConfig();
|
||||
if (!(id in current.mcpServers)) {
|
||||
// Idempotent DELETE — but still sweep the keychain in case a
|
||||
// prior delete failed after rewriting the file and orphaned
|
||||
// entries are sitting there.
|
||||
await secretStore.deleteAllForServer(id);
|
||||
return c.json({ ok: true });
|
||||
}
|
||||
delete current.mcpServers[id];
|
||||
await writeMcpAndTrackMtime(serializeStore(current));
|
||||
await secretStore.deleteAllForServer(id);
|
||||
return c.json({ ok: true });
|
||||
});
|
||||
} catch (error) {
|
||||
const keychainResp = keychainErrorResponse(c, error);
|
||||
if (keychainResp) return keychainResp;
|
||||
const msg = error instanceof Error ? error.message : String(error);
|
||||
return c.json({ error: `Failed to delete server: ${msg}` }, 500);
|
||||
}
|
||||
|
|
|
|||
|
|
@ -11,8 +11,13 @@ import type {
|
|||
MCPServerConfig,
|
||||
ServerEntry,
|
||||
ServerType,
|
||||
StdioServerConfig,
|
||||
StoredMCPServer,
|
||||
} from "./types.js";
|
||||
import {
|
||||
SECRET_FIELD_OAUTH_CLIENT_SECRET,
|
||||
envSecretField,
|
||||
} from "../auth/secret-fields.js";
|
||||
|
||||
// The full set of valid `type` discriminator values, used to reject anything
|
||||
// else read off disk so unknown strings can't propagate to narrowing sites.
|
||||
|
|
@ -274,6 +279,147 @@ export function serializeMcpConfig(entries: ServerEntry[]): string {
|
|||
return JSON.stringify(serverEntriesToMcpConfig(entries), null, 2);
|
||||
}
|
||||
|
||||
/**
|
||||
* Result of splitting secret values off a `StoredMCPServer` for keychain
|
||||
* persistence. `stripped` is what gets written to `mcp.json` on disk;
|
||||
* `secrets` is the (field → value) map the keychain backend writes.
|
||||
*
|
||||
* Empty-string values are not written to keychain (they have the same
|
||||
* semantic meaning as absence). Callers that want full reconcile
|
||||
* semantics on update should also call `deleteAllForServer` first.
|
||||
*/
|
||||
export interface ExtractedSecrets {
|
||||
stripped: StoredMCPServer;
|
||||
secrets: Record<string, string>;
|
||||
}
|
||||
|
||||
/**
|
||||
* Type guard for the stdio branch of `MCPServerConfig`. The `type` field
|
||||
* is optional on `StdioServerConfig` because stdio is the implicit
|
||||
* default for entries written without a `type` key (matches Claude
|
||||
* Desktop). So both `undefined` and the literal "stdio" route here.
|
||||
*/
|
||||
const isStdioStored = (
|
||||
stored: StoredMCPServer,
|
||||
): stored is StdioServerConfig & StoredMCPServer =>
|
||||
stored.type === "stdio" || stored.type === undefined;
|
||||
|
||||
/**
|
||||
* Strip secret values from a single on-disk entry. Returns the
|
||||
* sanitized disk shape (oauth.clientSecret removed; stdio env values
|
||||
* cleared to "") plus the map of field → value the keychain should
|
||||
* hold.
|
||||
*
|
||||
* stdio env keys are preserved with empty-string values rather than
|
||||
* dropped, so the on-disk file still documents the env interface the
|
||||
* server expects (a user reading mcp.json can see "this server uses
|
||||
* API_KEY and DB_PASSWORD" even though the values live in the
|
||||
* keychain). Round-tripping with another tool that reads mcp.json
|
||||
* gets the same key set but empty values, which is the intended
|
||||
* trade-off: the secret never reaches another tool, but the key list
|
||||
* is still visible.
|
||||
*/
|
||||
export function extractSecretsFromStored(
|
||||
stored: StoredMCPServer,
|
||||
): ExtractedSecrets {
|
||||
const secrets: Record<string, string> = {};
|
||||
const stripped: StoredMCPServer = { ...stored };
|
||||
|
||||
if (stored.oauth?.clientSecret) {
|
||||
secrets[SECRET_FIELD_OAUTH_CLIENT_SECRET] = stored.oauth.clientSecret;
|
||||
const restOauth: { clientId?: string; scopes?: string } = {};
|
||||
if (stored.oauth.clientId !== undefined)
|
||||
restOauth.clientId = stored.oauth.clientId;
|
||||
if (stored.oauth.scopes !== undefined)
|
||||
restOauth.scopes = stored.oauth.scopes;
|
||||
if (Object.keys(restOauth).length > 0) {
|
||||
stripped.oauth = restOauth;
|
||||
} else {
|
||||
delete (stripped as unknown as Record<string, unknown>).oauth;
|
||||
}
|
||||
}
|
||||
|
||||
if (isStdioStored(stripped)) {
|
||||
const env = stripped.env;
|
||||
if (env) {
|
||||
const newEnv: Record<string, string> = {};
|
||||
for (const [k, v] of Object.entries(env)) {
|
||||
if (typeof v === "string" && v.length > 0) {
|
||||
secrets[envSecretField(k)] = v;
|
||||
}
|
||||
newEnv[k] = "";
|
||||
}
|
||||
stripped.env = newEnv;
|
||||
}
|
||||
}
|
||||
|
||||
return { stripped, secrets };
|
||||
}
|
||||
|
||||
/**
|
||||
* Inverse of `extractSecretsFromStored`. Merges a pre-fetched secrets
|
||||
* record back into an on-disk entry. Used at the `/api/servers` GET
|
||||
* boundary so the browser receives the same effective shape it has
|
||||
* today.
|
||||
*
|
||||
* A missing key in `secrets` leaves the corresponding field alone,
|
||||
* which matters for stdio env: if the keychain doesn't have a value
|
||||
* for `env:KEY`, the on-disk empty string passes through unchanged.
|
||||
*/
|
||||
export function mergeSecretsIntoStored(
|
||||
stored: StoredMCPServer,
|
||||
secrets: Record<string, string>,
|
||||
): StoredMCPServer {
|
||||
const out: StoredMCPServer = { ...stored };
|
||||
|
||||
const oauthSecret = secrets[SECRET_FIELD_OAUTH_CLIENT_SECRET];
|
||||
if (oauthSecret) {
|
||||
out.oauth = { ...(out.oauth ?? {}), clientSecret: oauthSecret };
|
||||
}
|
||||
|
||||
if (isStdioStored(out) && out.env) {
|
||||
const newEnv: Record<string, string> = { ...out.env };
|
||||
let mutated = false;
|
||||
for (const k of Object.keys(out.env)) {
|
||||
const val = secrets[envSecretField(k)];
|
||||
if (val !== undefined) {
|
||||
newEnv[k] = val;
|
||||
mutated = true;
|
||||
}
|
||||
}
|
||||
if (mutated) {
|
||||
out.env = newEnv;
|
||||
}
|
||||
}
|
||||
|
||||
return out;
|
||||
}
|
||||
|
||||
/**
|
||||
* Enumerate the keychain field identifiers an on-disk entry expects
|
||||
* to find values for. Handlers use this to know which keychain entries
|
||||
* to fetch when rehydrating a server, and which keys to reconcile on
|
||||
* update (env keys removed by the user should drop their keychain
|
||||
* entries).
|
||||
*
|
||||
* Order is stable: OAuth first, then env keys in object iteration
|
||||
* order. Callers that diff old vs new field sets rely on stable
|
||||
* enumeration to avoid spurious churn.
|
||||
*/
|
||||
export function expectedSecretFields(stored: StoredMCPServer): string[] {
|
||||
const fields: string[] = [];
|
||||
// Always include OAuth slot — even if the entry has no `oauth` block
|
||||
// on disk, the keychain may hold a leftover entry from a prior
|
||||
// configuration that we want callers to be able to reconcile.
|
||||
fields.push(SECRET_FIELD_OAUTH_CLIENT_SECRET);
|
||||
if (isStdioStored(stored) && stored.env) {
|
||||
for (const k of Object.keys(stored.env)) {
|
||||
fields.push(envSecretField(k));
|
||||
}
|
||||
}
|
||||
return fields;
|
||||
}
|
||||
|
||||
/**
|
||||
* Default seeds written to `~/.mcp-inspector/mcp.json` on first launch when
|
||||
* the file is absent. Picked to cover the two shapes a developer reaches for
|
||||
|
|
|
|||
Loading…
Reference in a new issue