Local RAG MCP Server

Example Use Cases

• 1

  Augment LLM responses with up-to-date web context without sending data to third-party APIs.

• 2

  Perform live web searches and return concise, ranked context snippets (Markdown) to chat models.

• 3

  A locally runnable MCP server (Docker or uvx) to integrate retrieval-augmented generation into existing MCP-enabled clients.

Description

mcp-local-rag runs locally (or in Docker) as an MCP server to fetch recent web results, compute embeddings, rank relevance, and extract Markdown-ready context for language models. It uses DuckDuckGo for search and MediaPipe Text Embedder for embeddings, then returns top-k contextual snippets to the calling LLM. The tool is designed for privacy-conscious, API-free retrieval augmentation and integrates with MCP-capable clients like Claude Desktop, Cursor, and Goose.