PocketBase MCP Server (github.com)

🤖 AI Summary
PocketBase MCP Server is a Model Context Protocol (MCP) adapter that exposes full PocketBase functionality to AI assistants and other MCP clients, letting LLM-driven tools perform authentication, data and user management, collection/schema administration, and raw HTTP operations against any PocketBase API. It supports admin and regular user auth with session persistence (saveSession), auto-auth at startup via environment variables, multi-instance connections, and an SSE endpoint for live updates. You can run it via npm, configure via .env or pocketbase.config.json, and wire it into an MCP runner; default HTTP server is on port 3001. For AI/ML workflows this matters because it turns PocketBase into a first-class tool for agents — enabling programmatic creation/queries of collections and records, user lifecycle management, file uploads, and administrator operations directly from model-driven agents while preserving session state. It supports custom headers, filtering/sorting/pagination using PocketBase filter syntax, raw request passthroughs with multiple auth fallbacks (current session, env admin token, explicit token, or public), and consistent machine-readable error responses. A TOON output format option promises 30–60% token savings when interfacing with LLMs, making agent-to-database interactions more efficient and cost-effective for retrieval-augmented generation, autonomous agents, and test/CI pipelines.
Loading comments...
loading comments...