Skip to main content
The MCP server runs as a stdio subprocess spawned by your AI tool. Configure the API URL and Personal Access Token (PAT) once, then register the spawn command with Claude Code, Cursor, or another MCP client that supports stdio.
1

Generate a Personal Access Token

In the web app: Settings → Personal Access Tokens → Create. Copy the mr_pat_… value — it is shown once. Treat it like a password: it grants read/write access to all your resumes.
2

Configure the MCP server

This writes ~/.magic-resume/mcp.json. If you’re running the Core gateway locally, point --api-url at http://localhost:3110/api.
3

Register with your AI tool

This adds Kenmet to the user-scoped MCP list. Restart Claude Code so it picks up the new server.
4

Verify

In your AI tool, ask: “List my Kenmet resumes.” The model should call list_resumes and return your list. If it errors, run npx -y @magic-resume-kit/mcp config print and confirm the API URL and PAT are correct.
The MCP server communicates over stdio — it does not open a network port, listen on a socket, or use SSE. Your AI tool spawns it as a child process and shuts it down when the session ends.

How do I update or pin the MCP version?

Because the spawn command is npx -y @magic-resume-kit/mcp mcp, your AI tool pulls the latest version every time it starts. To pin a version explicitly:

How do I reset MCP credentials?

This removes ~/.magic-resume/mcp.json. Revoke the PAT in the web app too if it leaked.