openrecruiter-mcp

# openrecruiter-mcp Open Recruiter's recruiting tools, in any MCP client — Claude Desktop, Claude Code, Cursor, or anything else that speaks [MCP](https://modelcontextprotocol.io). Two audiences, from one package. A recruiter asks *"who in my pipeline fits this CUDA role?"* and their assistant ranks the candidates, reads the scores and explains the gaps. A consumer app asks on a job seeker's behalf — *"机器学习工程师, 东京"* — and gets job cards back, then submits a résumé to one of them. Two transports, too: **stdio** for a client on the same machine, **streamable HTTP** for a remote caller such as a chat backend. Nothing is uploaded anywhere except to the model provider you have already configured. ```bash pip install openrecruiter-mcp openrecruiter-mcp --help ``` ## Wire it into a client Add one entry to your client's `mcpServers` config. For Claude Desktop that file is `~/Library/Application Support/Claude/claude_desktop_config.json` on macOS, `%APPDATA%\Claude\claude_desktop_config.json` on Windows: ```json { "mcpServers": { "open-recruiter": { "command": "openrecruiter-mcp", "args": ["--data-dir", "/absolute/path/to/your/pipeline"], "env": { "ANTHROPIC_API_KEY": "sk-ant-...", "VOYAGE_API_KEY": "pa-..." } } } } ``` Nothing to clone and nothing to keep running — the client starts the process when it needs it. With [uv](https://docs.astral.sh/uv/) you can skip installing altogether: ```json { "command": "uvx", "args": ["openrecruiter-mcp", "--data-dir", "/absolute/path/to/your/pipeline"] } ``` <!-- readme-test: skip --> ```bash # Not released to PyPI yet? Install the same thing straight from the repository: uvx --from "git+https://github.com/miao4ai/open_recruiter.git#subdirectory=sdk/mcp" openrecruiter-mcp ``` ## Tools `--tools` picks the audience. They are separate sets because a job seeker's assistant has no business calling `set_candidate_status`. ### `--tools recruiter` (the default) Whatever the…