@jobtrack/mcp

A standalone Model Context Protocol server (stdio transport) for JobTrack — a personal job-application tracker with hybrid semantic search and duplicate detection. Point an MCP client (Claude Desktop, Claude Code) at it to read and write job applications directly: "log that I applied to Spotify today", "what's still awaiting a reply", "save this posting for later".

It talks to the SQLite database directly — the exact same services/*.service.ts functions the JobTrack REST API calls, via @jobtrack/api's exports map — so it doesn't need the API server running. Tools cover create/update/status-change for applications, companies, notes, tags and job openings, plus every read (list/get/search/dashboard) — deliberately no delete tools, so an MCP client can only add to or edit data, never destroy it.

Install

npm install -g @jobtrack/mcp

Then point an MCP client at the installed jobtrack-mcp command:

{
  "mcpServers": {
    "jobtrack": {
      "command": "jobtrack-mcp"
    }
  }
}

By default this reads/writes the same per-user database as the jobtrack tray app (%APPDATA%\jobtrack on Windows, ~/.local/share/jobtrack elsewhere) — install both and they share data with no extra configuration. Set JOBTRACK_HOME in the MCP client's env to point this at a different data directory instead.

Requires Node.js 24+ already installed on the machine — this runs the TypeScript sources via tsx rather than shipping a compiled, dependency-free binary.

Source

https://github.com/CuplexUser/JobTrack — see the main repo README's MCP server section for the full tool list and how it's implemented.