Document Search MCP Server

A local MCP server that provides read-only access to documents (Word, PDF, Excel, CSV, Visio, PNG, JPG) via the Model Context Protocol (MCP).

Features

• Read-only access: Enforced at the code level.

• Recursive scanning: Scans all subdirectories of configured directories.

• File type validation: Uses magic numbers to validate file types.

• Metadata extraction: Extracts metadata from all supported file types.

• Logging: Logs all actions to a file (configurable).

Related MCP server: Docalyze

Installation

1. Clone the repository:

   git clone https://github.com/bpweatherill/Document-Search-MCP-server.git
   cd Document-Search-MCP-server

2. Install dependencies:

   pip install -r requirements.txt

Usage

Start the Server

# Default: Scans ~/Documents
python -m src.main

# Custom directories and settings
python -m src.main --dir "C:\Users\Me\Documents" --dir "C:\Users\Me\Desktop" --max-file-size 50000000 --log-file "C:\logs\mcp.log" --log-level DEBUG

MCP Client Interaction

Send JSON-RPC requests over stdio. Example:

{
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "list_files",
    "arguments": {}
  }
}

Tools

Configuration

Security

• Read-only: All file operations are read-only.

• Path validation: Prevents directory traversal attacks.

• File type validation: Ensures files match their extensions.

Testing

Run unit tests:

python -m unittest discover tests