shinpr/mcp-local-rag

Local-first RAG server for developers. Semantic + keyword search for code and technical docs. Works with MCP or CLI. Fully private, zero setup.

What it solves

It provides a way to search through private local documents (PDF, DOCX, Markdown, and text files) without sending the data to an external embedding API. It allows AI assistants or users via a terminal to perform semantic and keyword searches on local files while keeping the data offline.

How it works

The project implements a local Retrieval-Augmented Generation (RAG) system using the Model Context Protocol (MCP). It indexes documents from a specified local directory, combining semantic similarity (via local embedding models) with keyword matching for technical terms. It can be integrated into MCP-compatible clients (like Claude Code, Cursor, or Codex) or used as a standalone CLI tool. It supports local model caching and can optionally use local vision models to generate captions for PDF figures.

Who it’s for

  • Developers and researchers who need to search private documentation offline.
  • Users of MCP-compatible AI clients who want to give their agents access to local knowledge bases.
  • People who prefer local-first AI tools to avoid API costs and privacy concerns.

Highlights

  • Fully Local: No API keys, Docker, or external databases are required; ingestion and search work offline after the initial model download.
  • Hybrid Search: Combines semantic intent with exact keyword matching for API names and error codes.
  • Broad Format Support: Indexes PDF, DOCX, Markdown, and text files.
  • MCP Integration: Seamlessly connects to AI agents via the Model Context Protocol.
  • Vision Capabilities: Optional local captioning for PDF figures to make them searchable.

Related

  • Project
    arabold/docs-mcp-server

    Grounded Docs MCP Server is an open‑source, locally‑run service that scrapes and indexes documentation (websites, GitHub, npm, PyPI, local files) for any library version, stores it (optionally with vector embeddings), and exposes a MCP‑compatible HTTP endpoint. AI coding assistants (Claude, Copilot, Gemini, etc.) can query the server to get exact, up‑to‑date answers, reducing hallucinations while keeping data private.

  • Dispatch
    Hugging Face MCP for Research: Connecting AI to Research Tools

    Hugging Face introduces the Research Tracker MCP, enabling AI agents to automate research discovery by integrating arXiv, GitHub, and Hugging Face through the Model Context Protocol.

  • Project
    blazickjp/arxiv-mcp-server

    A local MCP server for agent literature work. The differentiator is original-LaTeX section reads, BibTeX from arXiv metadata, and topic watches. Papers stay on disk. The working loop is paper ID → outline → one section → citations. Search is optional.

  • Dispatch
    Hugging Face Accelerating Document AI

    Hugging Face provides a comprehensive guide to using open-source multimodal models to automate document classification, parsing, and visual question answering for enterprise workflows.

  • Project
    nanbingxyz/5ire

    5ire is a free, cross‑provider AI chat assistant that uses the open Model‑Context‑Protocol (MCP) to attach tools (file system, databases, APIs, etc.) and includes a local multilingual vector store for RAG, a prompt library, bookmarks, and usage analytics.