OpenRoles runs an MCP server. An AI agent you connect to it works on your account as you: it reads your tracker and your resumes, logs and updates applications, saves resumes it writes, and scores them. It gets the plan you have and nothing more.

https://openroles.in/mcp

Step-by-step guides for each app are in the docs, under Connect your AI.

• claude.ai

  Settings → Connectors → Add custom connector, and paste the server address. Claude opens OpenRoles for you to sign in and allow it.

• ChatGPT

  Turn on developer mode under Settings → Connectors, then create a connector with the server address. You sign in to OpenRoles and allow it.

• Claude Code

  claude mcp add --transport http openroles https://openroles.in/mcp

  The first time it is used, it opens OpenRoles in your browser for you to sign in and allow it.

• Cursor and Claude Desktop

  Through mcp-remote, with a personal token in place of orl_pat_…. In Cursor this goes in ~/.cursor/mcp.json; in Claude Desktop, in its config file under Settings → Developer.

  { "mcpServers": { "openroles": { "command": "npx", "args": [ "-y", "mcp-remote", "https://openroles.in/mcp", "--header", "Authorization:${OPENROLES_AUTH}" ], "env": { "OPENROLES_AUTH": "Bearer orl_pat_…" } } } }

claude.ai, ChatGPT and Claude Code sign in through OpenRoles and need nothing else. Cursor, Claude Desktop and scripts of your own use a personal token. Make one under Account → Agents; it is shown once, and you can revoke it there at any time.

It will be able to:

• Read your tracker, your resumes and your saved answers (read)
• Log, update and delete applications, save your application answers, upload resume versions, rename and archive them (write)
• Score resumes, which counts against your plan the same as doing it here (ai)

It will not be able to edit a resume, change your plan, or pay for anything.

You can take this back any time under Account → Agents.