MCP Server
Connect Claude directly to your LitmusLab projects. Once connected, you can ask Claude things like "what's failing in the Checkout suite" or "trigger a run of the Nightly collection and tell me how it goes" — Claude calls the same endpoints the dashboard and Public API are built on, scoped to your own account.
Connect from Claude.ai
- In Claude.ai, go to Settings → Connectors → Add custom connector.
- Enter a name (e.g. "LitmusLab") and the server URL:
https://mcp.litmus-lab.com/mcp - In the Client ID field, enter:
Leave Client Secret blank. This server doesn't support Dynamic Client Registration, so the Client ID field is required and must be exactly this value — anything else, or leaving it blank, will fail to connect.claude-ai - Click Connect. You'll be redirected to log in to LitmusLab if you aren't already, then land on a consent screen asking whether to allow Claude access to your projects.
- Click Allow. You're redirected back to Claude.ai, connected.
No API key or secret to create — authentication happens through your normal LitmusLab login during that redirect.
Access & scope
A connected session acts as you — it sees exactly the projects you're a member of in the dashboard, nothing more. This is different from an API key, which grants access to every project an entire org owns. Two people connecting the same Claude.ai workspace to LitmusLab each get their own scoped access, tied to their own login.
Access is all-or-nothing for now — there's no read-only mode. Revoke access anytime from Settings → Connectors in Claude.ai, or by signing out other sessions from your LitmusLab account settings.
What it can do
| Category | Tools |
|---|---|
| Discovery | List your projects, a project's suites, API monitors, and UI monitors |
| Run history | List a suite's or collection's past runs; fetch one run's full results |
| Trigger & control | Trigger a suite or collection run (waits up to 50s for it to finish, then hands you a run ID to poll); request cooperative cancellation of a running run |
| Monitor data | List API monitor runs and captures, UI monitor captures and check runs; a composed overview (match rate, last capture time) for an API monitor |
| Test cases | List a suite's test cases |
Triggering a run doesn't block indefinitely — if it's still going after 50 seconds, you get back a run ID and a note to check on it, rather than the tool call hanging.
Troubleshooting
- "Connection issue" or reconnect prompts — usually means the previous access was revoked or expired. Disconnect and reconnect from Claude.ai's connector settings.
- Claude says it can't see a project — confirm you're actually a member of that project in LitmusLab (not just a member of the org that owns it) — the connector only sees your direct project access, same as the dashboard.
- Nothing happens after clicking Connect — check that you're logged into LitmusLab in the same browser; the consent redirect relies on your existing session.