Connect ChatGPT, Claude, Cursor, or another compatible hiring agent to Litmus over the Model Context Protocol.

MCP server

Litmus exposes a Model Context Protocol server at https://litmushiring.com/mcp/<your-org-id>/ so external agents — ChatGPT, Claude, Cursor, enterprise copilots, or your own automation — can discover roles, send assessment invites, read scores and reports, and record hiring decisions. Every request is scoped to the calling Clerk org via either OAuth (interactive clients) or an API key (headless). Find your org-scoped URL in Settings → API Keys in the dashboard.

Where to start

Quickstart — mint a key and make your first call in five minutes.
Authentication — OAuth setup for Claude Desktop / Code / Cursor and API keys for headless clients.
Tool reference — every shipped tool with arguments, return shapes, and a runnable Python example.
Troubleshooting — common failure modes and the fix for each.

What you can do today

Read — list and inspect submissions, pull pipeline counts and per-candidate status.
Operate — discover ordered role stages, send idempotent candidate invites, and record advance/reject decisions.
Build — create new assessments from your own files and iterate on existing ones with optimistic concurrency.
Audit — every MCP tool call lands in your dashboard's Audit Log tab so admins can see exactly what each credential did, when.

Assessment generation through the agent loop is on the roadmap but not yet exposed as an MCP tool; for now, build the assessment yourself and publish it via upload_assessment.