L

Local RAG

Local RAG for semantic document search without external API calls.

388 views
0 installs
Updated Nov 21, 2025
Not audited
Local RAG for semantic document search without external API calls.
  1. Determine the Folder for Your Documents
    Decide which folder on your computer contains the documents you want to search through. This folder will be referenced as BASE_DIR and the server will only have access to files within this directory.

  2. Open the FastMCP Connection Interface
    Go to the FastMCP connection interface and locate your installed “MCP Local RAG” server configuration.

  3. Set the BASE_DIR Environment Variable

    • In the configuration panel, find the field for ENV values.
    • Set BASE_DIR to the absolute path of your chosen documents folder (e.g., /Users/yourname/Documents/mydocs).
  4. (Optional) Set Advanced Configuration ENV Values
    You may also set the following optional ENV variables, depending on your needs:

    • DB_PATH: (default is ./lancedb) — Path to your LanceDB storage directory.
    • CACHE_DIR: (default is ./models) — Path to the model cache, if you want the model files stored elsewhere.
    • MODEL_NAME: (default is Xenova/all-MiniLM-L6-v2) — HuggingFace model ID, only change if you want a different Transformers.js-compatible embedding model.
    • MAX_FILE_SIZE, CHUNK_SIZE, CHUNK_OVERLAP: Adjust for performance or special needs. Defaults are usually sufficient.
  5. Save the Configuration
    Click to save/add the ENV values in the FastMCP interface.

  6. Restart the MCP Client

    • For Cursor: Fully quit and relaunch the app.
    • For Codex: Restart the IDE/extension.
    • For Claude Code: No restart needed; changes are picked up immediately.

No API keys or cloud credentials are needed. All data remains local.

You may now begin ingesting, searching, and managing documents using the configured RAG server via your AI client or FastMCP.

How to Install Local RAG

Install Local RAG MCP server with one click through FastMCP. Choose your preferred AI development tool below:

Claude Desktop

Click "Claude Desktop" in Quick Start

Cursor IDE

Click "Cursor IDE" in Quick Start

VS Code

Click "VS Code" in Quick Start

Alternatives to Local RAG

Looking for similar MCP servers? Browse other servers in the same categories on FastMCP, or check out the similar servers listed above.

Quick Start

View on GitHub

More for AI and Machine Learning

View All →

More for Developer Tools

View All →

Similar MCP Servers

Report Issue

Thank you! Your issue report has been submitted successfully.

Stay ahead of the MCP ecosystem

Get the top new MCP servers, trending tools, and dev tips delivered weekly. Free, no spam, unsubscribe anytime.

Join 2,847 developers. We send one email per week.