PocketBase MCP Server: Give AI Agents Access to Your Single-Binary Backend

Why Give AI Access to PocketBase?

PocketBase is a single-binary backend with embedded SQLite, real-time subscriptions, built-in auth, and file storage. When building features, you switch to the PocketBase Admin UI to check collection schemas, verify auth rules, test API filters, and browse uploaded files. Each context switch costs focus during frontend development.

An MCP server turns your PocketBase Admin UI into a set of structured tools that AI coding assistants (Claude, Cursor, Windsurf) can call directly. Instead of switching to the admin panel to check schemas or test queries, your AI reads them in real-time while you code.

What AI Agents Can Do with PocketBase Access

With a DataFaucet MCP server pointed at your PocketBase Admin, AI tools can:

• Query collections — list records with filters, expand relations, check field types and defaults
• Inspect auth rules — see collection-level list/view/create/update/delete rules with filter expressions
• Check API endpoints — verify which collections are exposed and their API rule configurations
• Browse files — list uploaded files, check storage paths, and verify file field configurations
• Test filters — run filtered queries against your data and inspect response shapes

Setting Up PocketBase as an MCP Server with DataFaucet

1. Open DataFaucet and start a capture session
2. Navigate to your PocketBase Admin UI (typically localhost:8090/_/)
3. Interact with collections, records, and settings — DataFaucet captures the API patterns
4. Deploy and connect to your AI coding tool

Related: Supabase MCP Server for Supabase, Directus MCP Server for Directus, Best MCP Servers for the full directory.

Example Use Cases

Building typed queries: Ask Claude "What fields does the posts collection have and what are the expand relations?" and get the full schema without opening the admin.

Debugging auth rules: Ask "What are the list and view rules on the orders collection?" instead of clicking through the Admin UI.

Frontend development: Ask "Show me the last 5 records in articles with their author expanded" and get live data while writing components.

Why DataFaucet for PocketBase

PocketBase keeps everything in a single binary with zero external dependencies, but navigating collection schemas, auth rules, and API configurations still requires the Admin UI. DataFaucet bridges that gap by giving AI tools direct access to your PocketBase instance through MCP, without sharing admin credentials or modifying your backend.

Works with local development instances, deployed PocketBase on Fly.io/Railway, or any environment running the binary.