* feat(mcp): require auth on /mcp when client is a plugin
Plugin hosts such as Claude Code only start OAuth for servers that 401 at
connect time. Matching Exa MCP, ?client=claude-code-plugin (any client
value containing "plugin") now gates /mcp the same way /mcp/oauth does,
while anonymous access on the public URL is unchanged.
Co-authored-by: Enes Gules <enesgules@users.noreply.github.com>
* chore(mcp): drop the SDK OAuth-helpers TODO
The v2 helpers (bearerAuthChallengeResponse, oauthMetadataResponse) assume
Bearer-only OAuth on a fetch() handler. This server also accepts API keys,
mixes anonymous and required routes, returns JSON-RPC 401 bodies, and
proxies authorization-server metadata live, so they are not a drop-in.
Co-authored-by: Enes Gules <enesgules@users.noreply.github.com>
* docs: keep the plugin client auth gate out of user-facing docs
The ?client=claude-code-plugin gate stays in the server and Claude plugin
URL. OAuth docs continue to describe /mcp/oauth only.
Co-authored-by: Enes Gules <enesgules@users.noreply.github.com>
* simplify Claude plugin auth tracking
* separate plugin and client metrics
* extract plugin request detection
* simplify MCP request handling
* extract authentication policy
* use OAuth for Claude Code plugin
* support API key or OAuth in Claude plugin
* simplify Claude plugin authentication
---------
Co-authored-by: Cursor Agent <cursoragent@cursor.com>
Co-authored-by: Enes Gules <enesgules@users.noreply.github.com>
Co-authored-by: Fahreddin Özcan <ozcanfahrettinn@gmail.com>
* fix(cli): refresh expired tokens for documentation commands
- Reuse getValidAccessToken in library and docs commands
- Avoid anonymous fallback when OAuth access tokens expire
* fix(cli): refresh expired tokens in skills suggest and generate
Route the remaining hand-rolled loadTokens/isTokenExpired checks through
getValidAccessToken so an expired token refreshes instead of silently
falling back to anonymous (skills suggest) or forcing a full re-login
(generate). Return undefined instead of null to match the optional
accessToken parameter on the API surface.
* fix(cli): preserve refresh_token and pin the auth wiring
RFC 6749 §6 permits a refresh response that omits refresh_token, in which
case the client keeps the one it holds. getValidAccessToken wrote the
response verbatim, dropping the stored token and silently logging the user
out at the next expiry. This PR widened that path from 2 commands to 6, so
fix it here.
Add a wiring test asserting each command passes a refreshed token to its
API call, and an eslint rule blocking loadTokens/isTokenExpired imports in
src/commands so the inline check cannot come back.
---------
Co-authored-by: Fahreddin Özcan <ozcanfahrettinn@gmail.com>
* fix(cli): write the API key as an Authorization header
Codex resolves a server's auth mode by checking only for
`bearer_token_env_var` or a header literally named `Authorization`
(`auth_status_before_discovery` in codex-rs/rmcp-client/src/auth_status.rs,
mirrored in `create_transport` in rmcp_client.rs). The custom
`CONTEXT7_API_KEY` header matched neither, so Codex fell through to any OAuth
credential stored for the same server name and URL and refreshed it during
startup. A dead refresh token then failed the server with `invalid_grant`
before the API key was ever sent, and re-running setup could not recover it
because setup writes config.toml and never touches the credential store.
The hosted endpoint accepts both header forms, so existing configs keep
working.
Two places keep the legacy header deliberately: the plugin .mcp.json files
default to `${CONTEXT7_API_KEY:-}`, and the server rejects `Bearer` with an
empty token while treating a missing header as anonymous; and `env` blocks in
stdio configs, where the name is an environment variable rather than a header.
* fix(plugins): send the API key via the Authorization header
The Claude and Copilot plugin configs default to `${CONTEXT7_API_KEY:-}`, and
both plugins document that an unset key still works over the anonymous tier.
The Bearer form cannot express that: the server rejects `Bearer` with an empty
token while treating an empty or missing Authorization header as anonymous.
The raw-key form satisfies both states. It is genuinely parsed rather than
ignored, verified by an invalid raw key being rejected, so a set key still
authenticates while an unset one falls back to anonymous as documented.
Once the server treats an empty-token Bearer as no header, these can move to
the `Bearer <key>` form used everywhere else.
* refactor(cli): narrow the Codex OAuth probe and trim its surface
Only `oauth` proves a stored credential exists. `not_logged_in` also covers
"no credential, server merely advertises OAuth", which is the normal state for
anyone who never logged in, so treating it as stale told most users their
config held a credential it did not.
Collapse the module to the two functions the call site needs, derive nothing
from a hand-maintained status list, and skip the subprocess entirely when the
server is not already in Codex's config. Drop the probe timeout to 1.5s and
kill with SIGKILL so it is a real ceiling rather than an intent, since the
result is only an advisory hint.
Lock the plugin manifests' raw-key form behind a test, so normalizing them to
`Bearer` for consistency with the CLI fails loudly instead of silently
breaking anonymous access.
* refactor(cli): drop the Codex OAuth cleanup note
The note existed because re-running setup could not rescue a stuck user. The
Authorization header change in this same branch makes it rescue them: Codex
never reads the stored credential once that header is present, so the
credential is inert and the hint only offered cosmetic cleanup.
Removing it drops a subprocess spawn from a user-facing path and a dependency
on the shape of `codex mcp get --json`, an external contract this repo does not
pin. The reason the header name matters moves to `withHeaders`, where the
decision is encoded.
* focus Context7 documentation queries
* narrow documentation query prompt changes
* remove focused from query prompts
* use lookup wording in query prompts
* distinguish documentation lookup from task
* allow live Pi test more time
* add prompt guidance changeset
The skill download step in `ctx7 setup` hits the git tree API on
api.github.com to enumerate a skill's files. When that host is blocked
or unreachable (while the docs host is fine), the fetch throws and setup
reports "Skill failed / fetch failed" (#2936).
Fall back to fetching the single SKILL.md directly from
raw.githubusercontent.com — the URL the docs API already resolves — so
single-file skills install even when api.github.com is not reachable.
* fix(cli): avoid shell for GitHub auth token
* fix(cli): document the shell-free constraint and harden gh token tests
Record why `gh auth token` must stay shell-free so the .cmd/.bat shim gap
is not "fixed" by re-adding `shell`, which would restore the cmd.exe
process that #2918 is about.
- reset mock implementations between tests so they stop leaking
- assert listSkillsFromGitHub's result; the tests passed green without it
- cover the GH_TOKEN fallback, which was previously untested
- reword the changeset: execSync spawned a shell on every platform, and
on Windows that shell was load-bearing rather than "unnecessary"
---------
Co-authored-by: Fahreddin Özcan <ozcanfahrettinn@gmail.com>
Fixes#2860
- Use npx ctx7@latest as the canonical CLI invocation in find-docs SKILL.md
- Add official library naming guidance matching rules/context7-cli.md
- De-emphasize global npm install as the primary workflow
- Add regression tests to keep skill and rule guidance aligned
Co-authored-by: syf2211 <syf2211@users.noreply.github.com>
* chore(deps): bump dependencies (combined dependabot updates)
Combines the safe dependabot dependency bumps into a single change:
- @modelcontextprotocol/sdk 1.25.2 -> 1.29.0 (mcp)
- undici 6.26.0 -> 8.3.0 (mcp)
- zod 4.3.5 -> 4.4.3 (mcp, tools-ai-sdk)
- commander 13.1.0 -> 15.0.0 (cli)
- ora 9.0.0 -> 9.4.0 (cli)
- dotenv 17.2.3 -> 17.4.2 (sdk, tools-ai-sdk, pi)
- @earendil-works/pi-coding-agent 0.75.5 -> 0.78.0 (pi)
eslint 9 -> 10 (#2703) is excluded: it is incompatible with the
pinned typescript-eslint v8 and breaks lint.
Verified: build, typecheck, lint, and tests pass.
* chore: add changesets for runtime dependency bumps
* fix(deps): pin undici to 7.x for Node 20 compatibility
undici 8 requires Node >=22.19.0 (it calls worker_threads.markAsUncloneable
unconditionally at module load), but CI and the release pipeline run Node 20,
which crashed the mcp test suite with 'markAsUncloneable is not a function'.
undici 7.27.0 guards that call and supports Node >=20.18.1.
* fix(sdk): avoid raw SyntaxError on non-JSON error responses
Wrap res.json() in the error path with .catch(() => ({})) so non-JSON
error bodies (HTML 502s, plain-text 429s, Cloudflare challenge pages)
fall through to res.statusText and always surface as a typed
Context7Error instead of a native SyntaxError.
Closes#1964
* chore: add changeset for sdk non-JSON error fix
* test(cli): pin home dir via HOME env instead of mocking os builtin
The storage-paths and auth-utils tests mocked the `os` module to fix
homedir, but that mock resolves inconsistently across Node versions and
worker pooling, leaking the real homedir on CI (/home/runner) and
failing 6 tests. os.homedir() reads $HOME first on POSIX, so stub HOME
(and clear XDG_* vars) for deterministic, order-independent paths with
no builtin-module mock. Also make the device-auth body assertion parse
client_id rather than matching the exact string, since hostname is
appended best-effort and varies by machine.
* fix(cli): recover library ID mangled by Git Bash on Windows
Git Bash rewrites a leading-slash argument like /facebook/react into a
Windows path under the Git install dir (C:/Program Files/Git/facebook/react),
so "ctx7 docs" rejected it as an invalid library ID. This mainly affected
users running ctx7 through Claude Code.
Detect and undo the conversion before validation, and point users at the
//owner/repo escape for install layouts that aren't auto-detected.
* chore: add changeset
* fix(cli): use XDG dirs for context files
* fix(cli): harden XDG migration and cover previews dir
- Move `generate` previews to $XDG_CACHE_HOME/context7/previews (was the
last writer recreating ~/.context7)
- Make legacy->XDG migration best-effort and fall back to reading the
legacy file so loadTokens/readUpdateState never throw or silently log out
- Split update-check read (legacy fallback) from write (always XDG target)
- Ignore relative/empty XDG_* values per the spec
- Fix non-hermetic XDG_STATE_HOME test that moved the real ~/.context7
cli-state into a temp dir; add storage-paths tests and a migration-failure
fallback test
* chore: add changeset for XDG directories
* fix(cli): enforce 0o600 on credentials after migration
rename preserves the legacy file's mode, so a credentials file that was
group/world-readable in ~/.context7 stayed readable after migrating to the
XDG path. chmod the target to 0o600 on migrate, and re-assert it after every
write (writeFileSync's mode is ignored when the file already exists).
---------
Co-authored-by: Fahreddin Özcan <ozcanfahrettinn@gmail.com>
The localhost-callback path is gone. Every install — laptop, SSH,
Codespace, Docker, CI — goes through the same boxed prompt and
verification page. Three reasons to make this the default:
- The localhost flow was broken anywhere the browser couldn't reach
127.0.0.1:52417 (SSH, Docker, Codespaces). Auto-detection via
SSH_CONNECTION / $DISPLAY was a half-fix that depended on env
vars users don't always set.
- Device flow works everywhere, has no random port-binding behavior,
and still ends in the same long-lived ctx7sk- API key.
- One UX path is simpler to support than two.
Drops the --device flag (it was the opt-in for what's now the
default). Older CLI versions (<= 0.5.0) continue to work against the
unchanged auth endpoints, so pinned installs are unaffected.
The legacy localhost machinery in utils/auth.ts is left in place
for now — nothing imports it from commands/auth.ts anymore, and a
follow-up can delete it once we're confident no rollback is needed.
* feat(cli): OAuth 2.0 device authorization flow
Adds RFC 8628 device-code login for headless / remote hosts (SSH,
Codespaces, Docker, CI) where the existing localhost-callback flow
can't work — the browser opens on the user's laptop while the
callback listener runs on the remote host, so the redirect target
is unreachable.
Device flow prints a verification URL and short code, then polls the
new /api/oauth/device/token endpoint. The user visits the URL on any
device, signs in, and approves; the CLI receives the same ctx7sk- API
key it would have gotten from the legacy flow.
- shouldUseDeviceFlow() auto-detects via SSH_CONNECTION /
SSH_CLIENT / SSH_TTY and missing $DISPLAY on Linux.
- ctx7 login --device forces it. ctx7 setup picks it up
automatically when resolveCliAuth needs to authenticate.
- pollDeviceToken returns a "transient" status for network errors
and 5xx responses so a flaky backend or Upstash blip doesn't end
the session — keeps polling until the device_code TTL elapses.
* polish(cli): boxed device-code prompt, Press-Enter, whoami success line
Tightens up the device-flow UX so it matches the patterns from gh /
stripe / wrangler:
- Wrap the user_code + verification URL in a boxen rounded box with
a title, gray border, and the code as the visual headline (green
bold, indented on its own line).
- Add a "Press Enter to open the browser, or Ctrl-C to quit..."
confirmation step in TTY mode so the user can read the code before
the browser steals focus. Skipped under --no-browser or non-TTY.
- Replace the generic "Login successful!" line with
"Logged in as <email> (<team>)" by fetching /api/dashboard/whoami
with the freshly minted token. Falls back to the old text if the
call fails.
- Tighten the mockShouldUseDeviceFlow signature in the test mock so
the spread-into-mock pattern typechecks.
No behavior change to the localhost-callback flow.
* test(cli): cover shouldUseDeviceFlow + start/poll + performDeviceLogin
auth-utils: SSH/$DISPLAY heuristics for shouldUseDeviceFlow; the
form-encoded start-device-authorization request shape and error
propagation; pollDeviceToken status mapping for each RFC 8628 code,
5xx -> transient, network error -> transient, and unknown 4xx ->
throw.
auth-commands: performDeviceLogin happy path (approved -> saveTokens
called), denied/expired return null and don't save, transient errors
keep polling instead of bailing, slow_down bumps the interval
(verified with fake timers), startDeviceAuthorization throwing exits
without polling, browser-open behavior under openBrowser=true/false.
Also covers the performLogin selector: forceDevice=true and
shouldUseDeviceFlow=true both route through performDeviceLogin
without hitting the localhost callback path.
28 new tests; suite is 235/235.
* docs(cli): tighten device-flow comments
Drop restatement; keep only WHY-bearing notes.
* fix(cli): default poll interval to 5s per RFC 8628 §3.2
The CLI was treating `interval` as required and would NaN-crash if
a future server omitted it. Spec requires clients to default to 5
when absent.
DeviceAuthorizationResponse.interval is now optional, and
performDeviceLogin uses DEFAULT_DEVICE_POLL_INTERVAL_SECONDS (5) as
the fallback. Test covers the missing-interval path.
* fix(cli): rfc 8628 spec gaps — backoff, hostname, bare verification_uri
Three small spec-compliance fixes from the §3.5 / §3.3 / §5.4 audit:
- §3.5: poll loop now bumps intervalMs by 5s on `transient` results
(network errors and 5xx) — the RFC requires unilateral backoff on
connection timeouts, and mirroring the slow_down handler is the
simplest correct response.
- §3.3: the boxed prompt now prints the bare verification_uri
alongside verification_uri_complete so screen readers / paper /
another device can still type the short form.
- §5.4: startDeviceAuthorization sends `os.hostname()` so the server
can show it on the verification page; the user can confirm the
device they're authorizing matches the one running the CLI.
Transient-backoff test rewritten with fake timers (the new +5s wait
made the old real-timer assertion blow past the 5s default timeout).
resolveMode treated --api-key as a non-interactive marker and
short-circuited to MCP, but --api-key is equally valid for CLI +
Skills mode (which authenticates skill downloads). Users who
preferred CLI mode were silently locked into MCP unless they
also passed --cli.
Remove options.apiKey from the auto-MCP OR-chain. --mcp / --cli /
--stdio / --oauth / -y still skip the prompt; --api-key alone now
falls through to the interactive mode picker.
* fix(cli): wire --antigravity and remove broken --universal in setup
The setup command advertised --universal and --antigravity flags but
neither was wired through getSelectedAgents, so passing them silently
fell back to auto-detection and wrote to the wrong directory (see #2695).
Remove --universal from setup entirely, and add a full Antigravity
SetupAgent config: skills under .agent/skills, MCP config at
~/.gemini/antigravity/mcp_config.json with serverUrl for HTTP, and
detection of .agent or ~/.gemini/antigravity.
* fix(cli): align antigravity setup with official Google docs
After verifying against Google Codelabs / Google Cloud Community
docs, correct the Antigravity config:
- MCP global path: ~/.gemini/config/mcp_config.json (Antigravity 2.0
shared config, replacing the older ~/.gemini/antigravity/ path
which an outdated github/github-mcp-server install guide cited).
- HTTP key: httpUrl (Gemini convention; antigravity is Gemini-based).
The previous serverUrl was sourced from the same outdated guide.
- Rule: append to GEMINI.md / ~/.gemini/GEMINI.md (Antigravity reads
Gemini-family rules, not a vendor-specific file).
- Project MCP: none documented; projectPaths is empty and setupAgent
/ remove falls back to globalPaths so --project --mcp still writes
to the correct location.
Skills stay at .agent/skills to keep the in-repo IDE_PATHS convention
consistent across setup, skill, and generate commands.
* chore(cli): move Antigravity to 5th in agent selection list
Match the natural ordering users expect in the checkbox prompt:
Claude Code, Cursor, OpenCode, Codex, Antigravity, Gemini CLI.
* fix(cli): antigravity HTTP key is serverUrl, not httpUrl
Antigravity rejects the entry with "serverURL or command must be
specified" when given httpUrl. Switch the HTTP entry back to
serverUrl (the github-mcp-server install guide had this right even
though its file path was outdated).
Also tighten the empty-projectPaths fallback in remove.ts: it
incorrectly leaked global state into project-scope detection,
making `remove --project` report Antigravity whenever a global
~/.gemini/config/mcp_config.json existed. Project-scope detect/
remove now no-ops for agents with no project-level MCP, while
setup still falls back to the global path so --project --mcp
--antigravity writes to the file Antigravity actually reads.
* fix(cli): wire --antigravity into the remove command
Symmetric to the setup fix: --antigravity was missing from
UninstallOptions and getSelectedAgents, so users had no CLI path
to undo a `ctx7 setup --antigravity` install.
* Add JSON output for skills list
* test(cli): resolve tempDir via realpath for macOS compatibility
On macOS, os.tmpdir() returns /var/folders/... but process.cwd()
reports the symlink-resolved /private/var/folders/... after chdir.
The JSON output uses the resolved cwd, so the test assertion mismatched
the unresolved tempDir on macOS.
* chore: add changeset for skills list --json
---------
Co-authored-by: Fahreddin Özcan <ozcanfahrettinn@gmail.com>
* fix(cli): declare @inquirer/core as direct dependency
selectOrInput.ts imports from @inquirer/core directly, but the package
was only resolvable as a transitive of @inquirer/prompts. Under pnpm's
isolated linker this fails with ERR_MODULE_NOT_FOUND at startup.
Fixes#2651
* chore: add changeset
* fix(cli): validate skill names to prevent path traversal on install
Adds boundary validation and containment checks so a remote SKILL.md
with a malicious name field (e.g. `name: ..`) cannot escape the skills
root during `ctx7 skills install`. Previously, the value flowed from
parseSkillFrontmatter to installSkillFiles unchecked, and the existing
traversal guard only verified files stayed inside the attacker-chosen
directory rather than the real skills root, enabling arbitrary file
writes outside `.claude/skills` (e.g. `.claude/settings.json` for
hook-driven RCE). symlinkSkill had the same trust issue and could
`rm(recursive: true)` arbitrary directories.
* chore: add changeset for skill name validation
* chore: soften changeset wording
readJsonConfig only caught file-read errors, not JSON.parse errors.
During `ctx7 remove`, the detector iterates every agent's well-known
config path; an unparseable JSON file at any of them (e.g. a
hand-edited ~/.claude.json) crashed the command with an unhandled
SyntaxError before it could do anything.
Wrap the readJsonConfig call in hasMcpConfig with a try/catch that
logs a warning naming the path and parse error and skips that agent.
Keep readJsonConfig itself strict so write paths in setup and
uninstallMcp continue to surface failures via their existing handling.
github-actions[bot]