About

This project provides an MCP server for 0xPdf, enabling schema-first PDF to structured JSON conversion specifically designed for AI agents. It acts as a wrapper for the 0xPdf HTTP API, preserving its authentication, billing, and rate limiting features. AI agents can leverage this server to parse invoices and forms into a defined JSON schema, generate schemas dynamically, and manage asynchronous parsing jobs.

Key Features

• Schema-First Parsing: Converts PDF documents into structured JSON based on a provided schema.
• AI Agent Integration: Designed to work seamlessly with AI agents like Claude and Cursor.
• 0xPdf API Wrapper: Utilizes the robust 0xPdf HTTP API for core functionality.
• Schema Generation: Supports generating JSON schemas from natural language descriptions using AI.
• Asynchronous Job Processing: Allows agents to submit parsing jobs and poll for completion.
• Tool Support: Includes tools for synchronous and streaming PDF parsing, job management, schema handling, and pricing information.

Installation

Install via pip:

pip install oxpdf-mcp

Or using uvx:

uvx oxpdf-mcp

An API key from 0xpdf.io/dashboard is required.

Configuration for Cursor/Claude Desktop

Configure your MCP settings with the provided JSON structure, including API endpoints, keys, and allowed hosts/file roots.

Environment Variables

Key environment variables include:

• PDF_PARSING_API_KEY: Your API key (required).
• PDF_PARSING_API_BASE_URL: The base URL for the 0xPdf API.
• PDF_PARSING_ALLOWED_API_HOSTS: Hosts allowed for API calls (e.g., api.0xpdf.io).
• PDF_PARSING_REQUIRE_HTTPS: Set to true for production environments.
• PDF_PARSING_ALLOWED_FILE_ROOTS: Specifies allowed directories for PDF files.

SDKs

In addition to the MCP server, official SDKs are available:

• Python: oxpdf
• JavaScript/TypeScript: @0xpdf/client

License

This project is licensed under the MIT License.