Tika MCP Server

This project provides a Model Context Protocol (MCP) server for extracting content and metadata from files using Apache Tika.

Overview

The Tika MCP server allows AI assistants to extract text and metadata from various file formats (PDF, DOCX, images with OCR, etc.) using Apache Tika. This enables AI assistants to understand and work with the content of files that users upload.

Features

• Extract text content from various file formats

• Extract metadata (author, creation date, etc.) from files

• Support for PDF, DOCX, images, and many other formats

• Simple JSON-RPC API following the Model Context Protocol

Requirements

• Python 3.6+

• Apache Tika server running (default: http://localhost:9998)

• MCP-compatible client

Installation

1. Clone this repository

2. Install dependencies: pip install -r requirements.txt

3. Register the MCP server: python -m app.register_mcp_server

Usage

The Tika MCP server provides a single tool:

extract_file

Extracts content and metadata from a file using Apache Tika.

Parameters:

• file_path: Path to the file to extract content from

• tika_url: URL of the running Tika server (default: http://localhost:9998)

Returns:

• metadata: Dictionary of metadata extracted from the file

• content: Array of content blocks extracted from the file

Testing

Several test scripts are provided to verify the functionality:

• app/test_tika_simple.py: Tests the Tika client directly

• app/test_simple_mcp.py: Tests the MCP server using the JSON-RPC protocol

Project Structure

• app/: Main application code

  • simple_mcp_server.py: MCP server implementation

  • tika_client.py: Client for Apache Tika

  • model.py: Data models and business logic

  • register_mcp_server.py: Script to register the MCP server

• examples/: Example files for testing

• requirements.txt: Python dependencies

Setup

Get a venv using either:

uv venv

or

python3 -m venv .venv

Activate the virtual environment and install dependencies:

source .venv/bin/activate
pip install -r requirements.txt

Running the MCP Server

Start the Apache Tika server (if not already running):

docker run -d -p 9998:9998 apache/tika

Register and run the MCP server:

python -m app.register_mcp_server

License

MIT