MCP connector
The TiLab MCP connector exposes jobs, applications, candidates and pipeline data through the Model Context Protocol. Claude can search and read your hiring records through it, and act on them only when you grant Read + write access.
At a glance
Plans
The MCP connector is part of the paid Basic and Growth plans, both coming soon. Every call is checked against the connected user's live workspace membership and role. This page does not promise a specific AI client's behaviour; that client's own terms govern what it does with the data it reads.
How a connection works
The connector follows the protocol's standard discovery and OAuth flow, so Claude can find it and ask for access without a custom integration. Nothing is shared until a signed-in user approves the request.
Add TiLab's MCP endpoint to Claude as a custom connector. Claude discovers the authorisation server from the published well-known metadata, and on the consent screen you choose which workspace the session connects to.
Sign in to TiLab and review the request. Choose Read to let the client search and view records, or Read + write to also let it make changes recorded as you.
The client calls the tools it needs: list open jobs, search candidates, read a CV, or pull pipeline analytics for a role.
Adding a note, updating a candidate, re-running CV parsing or moving an application to another stage requires Read + write. With Read access, those tools refuse the call.
Active connector sessions are listed under Settings → Connectors, where a session owner or workspace admin can revoke them.
What the tools cover
The tools map to the same records your team already uses in the workspace. They do not create a parallel data store, and every call is resolved against the connected user's role and workspace.
List and open jobs, list applications for a role, and read a single application with its stage and context.
Search candidates, open a profile, and read the text of an uploaded CV.
Add a note, update candidate fields, move an application to another pipeline stage, or re-run CV parsing when a file was read badly.
Pull aggregate pipeline analytics for a job and read the workspace's own configuration, so answers reflect how your team actually works.
Before you connect
MCP makes the ATS reachable from the tools your recruiters already use. That is useful precisely because it is real access, so decide the scope, the person and the review step before switching it on.
Read-heavy work where a recruiter checks the answer before anything changes.
Actions that change a candidate's position or record, and anything that amounts to a hiring decision.
Common questions
Claude, as a custom connector on claude.ai, the desktop and mobile apps, and Claude Code. Other hosted AI clients cannot complete sign-in today. Contact the team if you use a different client.
Basic and Growth, the paid plans, which are coming soon. Contact the team to talk through how you want to use it.
Admins and Recruiters. Access is resolved against the signed-in user's live membership each time, so removing someone from the workspace removes their MCP access too.
Only if a user granted Read + write. Read sessions cannot call the tools that add notes, update candidates, re-run CV parsing or move stages.
Explore the workflow
The MCP connector comes with Basic and Growth. Tell us what you want Claude to do in your hiring workflow, and we will help you start with Read access.
Contact the team