Tika MCP Extractor Server

Overview

The Tika MCP Extractor Server is a Model Context Protocol (MCP) compliant server that uses Apache Tika to extract content and metadata from files in various formats (e.g., PDF, DOCX, TXT, HTML, images) stored in a files-to-extract directory. It supports conversion to HTML (with optional CSS styling for better readability) or plain text and provides tools to list files and retrieve metadata. Built with Java 23, Spring Boot, Jetty, and the MCP SDK (0.11.0), it integrates with MCP-compliant clients like Claude Desktop or MCP Inspector.

The server exposes four MCP tools:

• extract-to-html: Converts file content to HTML (with embedded CSS).
• extract-text: Extracts plain text.
• list-available-files: Lists files in the directory with details.
• get-file-metadata: Retrieves detailed file metadata.

It also provides REST endpoints for testing, including a new endpoint to serve raw HTML directly for browser rendering. All operations are local, requiring no internet access, making it ideal for secure document processing workflows.

Features

• File Extraction: Converts file content to HTML (with CSS for readability) or plain text using Apache Tika.
• Metadata Extraction: Retrieves metadata like title, author, content type, and creation date.
• File Listing: Scans files-to-extract for files, providing size, MIME type, and modification details.
• MCP Integration: Four synchronous tools with JSON schema validation.
• REST Testing Endpoints:
  • GET /api/test/list: Lists available files.
  • POST /api/test/extract-html: Extracts file content as JSON with HTML string.
  • POST /api/test/extract-text: Extracts file content as plain text in JSON.
  • POST /api/test/raw-html: Serves raw HTML directly (renderable in browsers).
  • GET/POST /api/health: Checks server and directory status.
• CORS Support: Enabled for all REST endpoints for web-based testing.
• Configurability: Settings (port, directory, Tika options) via application.properties.
• Error Handling: Robust checks for file existence, readability, and parsing errors.
• Logging: Console logs with DEBUG support for Tika and PDFBox.

Prerequisites

• Java: JDK 23+ (tested with OpenJDK 24.0.2).
• Maven: Version 3.6+ for dependency management and building.
• Supported File Formats: PDF, DOCX, TXT, HTML, images, etc., handled by Apache Tika 2.9.1 and PDFBox 2.0.29.
• Optional: IntelliJ IDEA for development (output indicates IntelliJ usage, but any IDE or CLI works).
• Local Files: Place files in files-to-extract directory; no internet required.

Installation

1. Clone the Repository (if hosted):

   git clone https://github.com/RayenMalouche/MCP-PDF-Extractor-server.git
   cd MCP-PDF-Extractor-server
   

2. Create the Files Directory:

• The server reads from files-to-extract (configurable).
• Create it:

  mkdir files-to-extract
  

• Add sample files (e.g., sample.pdf, document.docx) for testing.

3. Build the Project:

• Use Maven to compile and resolve dependencies:

  mvn clean install
  

• Outputs executable JAR in target/.

Configuration

Settings are defined in src/main/resources/application.properties:

# Tika MCP Extractor Server Configuration
spring.application.name=TikaExtractorMCPServer

# Server Configuration
server.port=45453

# Tika Configuration
tika.max.string.length=-1
tika.detect.language=false

# File Processing Configuration
files.directory=files-to-extract
files.max.size=52428800

# Logging Configuration
logging.level.org.apache.tika=DEBUG
logging.level.org.apache.pdfbox=DEBUG

• spring.application.name: Application name for Spring Boot.
• server.port: HTTP port (default: 45453).
• tika.max.string.length: Sets max string length for Tika (-1 = unlimited).
• tika.detect.language: Disables langu

... View full README on GitHub