MCP Server

What is MCP Server?

MCP Server is an implementation of the Model Context Protocol (MCP) using Node.js and TypeScript. It provides document management, search, and question-answering capabilities for developers building AI applications that need to retrieve and interact with document content via MCP-compatible clients.

How to use MCP Server?

Clone the repository, install dependencies with npm install, and build the project with npm run build. Start the server using npm start for stdio transport or npm start -- --transport=sse for SSE transport. Configuration is done via environment variables such as PORT and DOCUMENT_DIRECTORIES.

Key features of MCP Server

• Cross-document search tool for retrieving relevant information
• Document management with listing and metadata resources
• Document Q&A prompt template
• Support for stdio and SSE transport methods
• Comprehensive logging system
• Readiness management and graceful shutdown

Use cases of MCP Server

• Building AI assistants that retrieve information from a document archive
• Implementing document Q&A workflows in MCP-compatible chat interfaces
• Integrating document search into automated knowledge retrieval pipelines

FAQ from MCP Server

What are the runtime requirements for MCP Server?

Node.js 22 or later is required.

Which transport methods does MCP Server support?

The server supports stdio for direct integration with MCP clients and SSE for web-based usage.

How can I configure MCP Server?

Set environment variables such as PORT, LOG_LEVEL, DOCUMENT_DIRECTORIES, EMBEDDING_MODEL, and LANCEDB_PATH.

What default embedding model does MCP Server use?

The default embedding model is sentence-transformers/all-MiniLM-L6-v2.

Where is the vector database stored by default?

By default, LanceDB stores data in the ./vectordb directory.