Verified

Local Rag MCP Server

Private document search using local indexing and hybrid retrieval methods.

Data last scanned today · Reviewed today

20 1 dangerous tool10 tools★ 387TypeScriptMIT

Overview

Local RAG allows users to search across PDF, DOCX, Markdown, and text files directly on their machine. It bypasses external embedding APIs, ensuring that all processing, indexing, and storage remains local after the initial model download.

Our verdict

This server carries a high-risk score of 45/100, primarily due to the fact that its authentication methods and read-only status have not yet been editorially reviewed. Given that it interacts directly with your local file system, users should exercise caution until further audits confirm the safety of its file operations. It is best suited for power users who are comfortable managing local dependencies and require offline, private document retrieval capabilities.

Supports hybrid search with semantic and keyword matching
Uses semantic chunking to preserve code block integrity
Small footprint with only nine direct dependencies
Fully functional offline after the initial model download

Security status and file-read safety remain unverified
Maintained by community contributors rather than an official vendor

Setup

Claude Desktop

{
  "mcpServers": {
    "mcp-local-rag": {
      "args": [
        "-y",
        "mcp-local-rag"
      ],
      "command": "npx"
    }
  }
}

Paste into ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows), then restart Claude Desktop. Some servers require API keys or tokens as environment variables — check the project's README and add an "env" object if needed.

No read-only mode — running this grants full read/write access.
No built-in authentication — anyone with access to this config can invoke every tool it exposes.

Tools

Tool	Description	Risk
sync_start	Reconcile the index with all configured roots or one path	medium
sync_status	Poll a running sync job	low
ingest_file	Ingest or replace one file	medium
ingest_data	Ingest text, Markdown, or HTML already held by the client	medium
query_documents	Search with semantic matching and keyword boost	low
read_chunk_neighbors	Read surrounding chunks from a search result	low
list_files	Show supported files and their ingestion state	low
delete_file	Delete an indexed file or an ingest_data item	high
status	Show index and search status	low
List all ingested files	List all files that have been processed and ingested into the index.	low

Compatibility

Client	Local	Docker	Remote	Read-only
ChatGPT
Claude Desktop
Cursor
VS Code
Windsurf

Frequently asked questions

›What are the system requirements for installing Local RAG?

You need Node.js 22 or later and a directory containing your documents. You must have internet access during the first run to download the npm package and the embedding model.

›How do I configure the search directory?

Set the BASE_DIR environment variable to the absolute path of your document directory. This directory also serves as the security boundary for all file operations.

›What file formats are supported for ingestion?

The server supports PDF, DOCX, Markdown, and text files for direct directory syncing. It also accepts text, Markdown, or HTML passed from an MCP client via the ingest_data tool.

›Are source code files or spreadsheets supported?

No, Excel, PowerPoint, standalone images, and source-code file extensions are not supported by the file ingestion process.

›How does the search functionality work?

The server uses a hybrid search approach that combines semantic similarity with keyword matching to ensure both conceptual intent and exact technical terms are retrieved.

›Is internet access required after the initial setup?

No, after the initial model download, document ingestion and search operations function entirely offline.