Context Book

What is Context Book?

Context Book is an MCP server that gives AI tools a persistent, searchable knowledge library, so they don't start from scratch on every conversation. It uses PostgreSQL with pgvector and pg_trgm for storage and Voyage AI for embeddings. The system consists of two Go binaries: an API server for authentication, book/page CRUD, and a dashboard; and an MCP server exposing eight Bearer-authenticated tools for AI agents.

How to use Context Book?

Install prerequisites: Go 1.26+, Node.js 22+, PostgreSQL 16+ with pgvector and pg_trgm extensions, and a Voyage AI API key. Set up the database, configure environment variables, then run go run ./cmd/api/main.go and go run ./cmd/mcp/main.go from the backend directory. Connect any MCP-compatible client by pointing it to http://localhost:8081/mcp, optionally run the frontend with npm run dev in the frontend directory. For Cursor, add the server URL to .cursor/mcp.json.

Key features of Context Book

• Eight MCP tools: create/update/list/get books, insert/update/delete/search pages, and a readme tool.
• Semantic search across all books using Voyage AI embeddings.
• Automatic embedding generation on page insert or update.
• OAuth 2.0 PKCE for the API server; Bearer token authentication for MCP tools.
• Two separate servers: API (REST, port 8080) and MCP (port 8081).
• Optional React frontend dashboard for managing books and pages.

Use cases of Context Book

• Provide AI agents with a persistent knowledge base that survives conversation resets.
• Avoid re-explaining project context, guidelines, or documentation to tools like Claude, Cursor, or Windsurf.
• Organize information into named “books” with ordered “pages” for structured recall.
• Perform semantic searches across all stored content to retrieve relevant context on demand.

FAQ from Context Book

What are the runtime dependencies?

Go 1.26+, Node.js 22+, PostgreSQL 16+ with pgvector and pg_trgm extensions, and a Voyage AI API key.

How do I connect an AI client?

Point the MCP-compatible client to http://localhost:8081/mcp. For Cursor, add the server URL to .cursor/mcp.json as shown in the quick start.

What authentication is used?

The API server uses OAuth 2.0 PKCE. The MCP server requires a Bearer token per request, scoped to the authenticated user.

Where does data live and how is it indexed?

Data is stored in PostgreSQL with pgvector for semantic search and pg_trgm for text matching. Migrations run automatically on API server startup.

What transports and ports are used?

The API server listens on :8080 for REST and the MCP server on :8081 for the MCP protocol over HTTP.