trexi is headless by design — the product surface is the API and tooling rather than a heavy UI. That makes it a natural fit for Claude Code and other agents via the Model Context Protocol (MCP).
This page describes the intended integration model. As the CLI and MCP server surface stabilizes, concrete tool names and setup steps will be documented here.

​
The idea

Instead of clicking through a dashboard, you describe hiring tasks to an agent:
  • Post or update a job and open it on the public board.
  • Triage incoming applications — move candidates from new → reviewing.
  • Summarize a candidate pool for a given role.
  • Reject or advance applications in bulk based on criteria you specify.
The agent calls the backend API (the same surface the dashboard uses), so anything the UI can do is scriptable.

​
Where things live

  • The backend (apps/backend) owns ATS logic and the integration points.
  • Auth flows through Clerk; an agent acts within an organization’s context.
  • The public board ({slug}.trexi.ai) reflects whatever jobs are marked open.
See Architecture for the full picture and Jobs & Applications for the domain model an agent operates on.