mcp-proxyml

# mcp-proxyml MCP server for the [ProxyML](https://proxyml.ai) API. Gives Claude (and other MCP clients) direct access to ProxyML's surrogate modelling and explainability tools. ## Prerequisites - A ProxyML API key — sign up at [proxyml.ai](https://proxyml.ai) - [uv](https://docs.astral.sh/uv/getting-started/installation/) (required for `uvx` installation method) ```bash curl -LsSf https://astral.sh/uv/install.sh | sh ``` Or if you use `pip`: ```pip install uv``` ## Installation ### Claude Desktop Add to your `claude_desktop_config.json`: ```json { "mcpServers": { "proxyml": { "command": "uvx", "args": ["mcp-proxyml"], "env": { "PROXYML_API_KEY": "your-api-key-here" } } } } ``` The config file is at: - **macOS**: `~/Library/Application Support/Claude/claude_desktop_config.json` - **Windows**: `%APPDATA%\Claude\claude_desktop_config.json` Then restart Claude Desktop. ### Claude Code ```bash claude mcp add-json proxyml '{"command":"uvx","args":["mcp-proxyml"],"env":{"PROXYML_API_KEY":"your-api-key-here"}}' ``` ### Other MCP clients ```bash pip install mcp-proxyml PROXYML_API_KEY=your-key mcp-proxyml ``` ### Local challenger training (optional) `proxyml_train_challenger` and `proxyml_score_champion` train a small model locally via the `proxyml` SDK and require the `challenger` extra, which pulls in scikit-learn and scipy: ```bash pip install 'mcp-proxyml[challenger]' ``` Without it, those two tools return a clear error telling you to install the extra rather than failing to start the server. ## Environment variables | Variable | Required | Description | |---|---|---| | `PROXYML_API_KEY` | Yes | Your ProxyML API key | ## Security & permissions ### What this server can access **Filesystem** — `proxyml_infer_schema` and `proxyml_train_challenger` each read one CSV file at the path you provide. No other tool reads or writes local files. **Network** — outbound HTTPS only, to `https://api.proxyml.ai` (o…