Claude-Patent-Creator provides an MCP server interface for programmatic patent analysis, generation, and search workflows. It exposes a suite of MCP tools for MPEP and patent search, prior art discovery, claims and specification reviews, and diagram generation, all accessible via MCP clients (including Claude Code integrations or other MCP clients). Typical usage involves running the Python-based server, then calling its MCP endpoints to perform hybrid RAG patent searches, review claims for compliance, generate diagrams, and assemble patent drafts. The server pairs with Claude Code Plugin capabilities, including 15 skills and 10 autonomous agents, to automate long-running workflows like comprehensive patent creation, prior art research, and full reviews through slash commands and event-driven hooks.

Prerequisites:

• Python 3.9+ installed on your system
• Git (optional, for cloning)
• MCP client setup if you plan to connect from an external client

Installation steps:

1. Clone the repository (if you have the repo):

   git clone https://github.com/RobThePCGuy/Claude-Patent-Creator.git cd Claude-Patent-Creator

2. Create and activate a virtual environment (recommended):

   python -m venv venv

   Linux/macOS

   source venv/bin/activate

   Windows

   venv\Scripts\activate

3. Install the package in editable mode (as per README guidance):

   pip install -e .

4. Run the setup to initialize MCP server integration:

   patent-creator setup

5. Run the MCP server via the CLI entry point (example):

   patent-creator health

   or start a background MCP server if supported by your setup

Prerequisites noted in the project include optional GPU support, BigQuery access, and Graphviz for diagram generation. If you intend to leverage GPU-accelerated indexing, ensure CUDA is installed and a compatible PyTorch version is selected during setup.

Tips and common issues:

• Ensure Python 3.9+ is consistently used across environments to avoid dependency conflicts.
• If you encounter GPU-related issues, verify CUDA 12.8 compatibility and that the correct PyTorch variant is installed during setup.
• For diagram generation, Graphviz must be installed at the system level (not just as a Python package).
• BigQuery access requires proper Google Cloud credentials; verify authentication with google-cloud-bigquery before running patent searches.
• The MCP server relies on the patent-creator CLI; use patent-creator commands to manage setup, health checks, and index rebuilding as needed.
• If replacing the CLI name, keep the same module invocation in the MCP config (python -m patent_creator) to maintain compatibility with the server bootstrap.
• When running in standalone Claude Code mode, ensure the plugin and setup wizard steps are completed to enable full functionality.