Shared Knowledge MCP

Example Use Cases

• 1

  Search and retrieve authoritative passages from a shared project knowledge base (Markdown and text files) with contextual metadata and location info.

• 2

  Consistent RAG-powered responses across multiple assistant platforms (CLINE, Cursor, Windsurf, Claude Desktop) using the same indexed knowledge.

• 3

  Flexible vector-store backends (local HNSWLib or managed services like Pinecone/Weaviate) to match deployment and scale requirements.

Description

This TypeScript-based MCP server ingests Markdown and text files, splits them into chunks, vectors them with the OpenAI API, and serves similarity search results to multiple AI assistants (CLINE, Cursor, Windsurf, Claude Desktop). It supports multiple vector stores (HNSWLib, Chroma, Pinecone, Milvus, Weaviate) via an abstracted interface, enabling easy switching and scalability. The server returns rich search results including location, document type, summaries, keywords, and relevance explanations to help LLMs use retrieved context effectively.