USPTO Patent MCP Server by riemannzeta - MCP Server

USPTO Patent MCP Server

A FastMCP server for accessing United States Patent and Trademark Office (USPTO) patent and patent application data through the Patent Public Search API, the Open Data Portal (ODP) API, and Google Patents Public Datasets via BigQuery. Using this server, Claude Desktop can pull data from the USPTO APIs or search through 90M+ patent publications from 17+ countries via Google's BigQuery:

For an introduction to MCP servers see Introducing the Model Context Protcol.

Special thanks to Parker Hancock, author of the amazing Patent Client project, for blazing the trail to understanding of the string of requests and responses needed to pull data through the Public Search API.

Features

This server provides tools for:

Patent Search - Search for patents and patent applications across USPTO and Google Patents databases
Full Text Documents - Get complete text of patents including claims, description, etc.
PDF Downloads - Download patents as PDF files (Claude Desktop doesn't support this as a client currently)
Metadata - Access patent bibliographic information, assignments, and litigation data
Google Patents Integration - Access 90M+ patent publications from 17+ countries via BigQuery
Advanced Search - Search by inventor, assignee, CPC classification, and more

API Sources

This server interacts with three patent data sources:

ppubs.uspto.gov - For full text document access, PDF downloads, and advanced search
api.uspto.gov - For metadata, continuity information, transactions, and assignments
Google Patents Public Datasets (BigQuery) - For comprehensive patent search across 90M+ publications from 17+ countries

Prerequisites

Python 3.10-3.13 (3.12 recommended)
Claude Desktop (for integration). Other models and MCP clients have not been tested.
For Patent Public Search requests, no API Key is required, but there are rate limits. This API is not meant for bulk downloads.
For ODP API requests, a USPTO ODP API Key (see below).
For Google Patents: A Google Cloud account with BigQuery API enabled (see Google Cloud Setup below).
UV for python version and dependency management.

If you're a python developer, but still unfamiliar with uv, you're in for a treat. It's faster and easier than having a separate python version manager (like pyenv) and setting up, activating, and maintaining virtual environments with venv and pip.

If you don't already have uv installed, curl -LsSf https://astral.sh/uv/install.sh | sh should do the trick.

Installation

Clone this repository:

git clone https://github.com/riemannzeta/patent_mcp_server
cd patent_mcp_server

Install dependencies with uv:
```
uv sync
```
If installed correctly, then:
```
uv run patent-mcp-server
```
Should write:
```
INFO     Starting USPTO Patent MCP server with stdio transport
```
to the console. With an API key installed in the environment and Claude Desktop configured, the patents MCP server is ready.

API Key Setup

USPTO API Key

To use the api.uspto.gov tools, you need to obtain an Open Data Portal (ODP) API key:

Visit USPTO's Getting Started page and follow the instructions to request an API key if you don't already have one.
Create a .env file in the patent_mcp_server directory with your API key:
```
USPTO_API_KEY=your_actual_key_here
```
You don't need quotes around your key. The ppubs tools will run without this API key, but the API key is required for the api.uspto.gov tools.

Google Cloud Setup (for Google Patents)

To use Google Patents Public Datasets, you need to set up Google Cloud credentials:

Create a Google Cloud Project:
- Go to Google Cloud Console
- Create a new project or select an existing one
- Note your Project ID
Enable BigQuery API:
- In your project, go to "APIs & Services" > "Library"
- Search for "BigQuery API" and enable it
Create Service Account Credentials:
- Go to "APIs & Services" > "Credentials"
- Click "Create Credentials" > "Service Account"
- Give it a name (e.g., "patent-mcp-bigquery")
- Grant it the "BigQuery User" role
- Click "Done"
- Click on the created service account
- Go to "Keys" tab > "Add Key" > "Create new key"
- Choose JSON format and download the key file

Configure Environment Variables: Add to your .env file:

GOOGLE_CLOUD_PROJECT=your-project-id
GOOGLE_APPLICATION_CREDENTIALS=/path/to/your/service-account-key.json

Cost Information: Google BigQuery provides 1TB of free queries per month. After that, queries cost $5 per TB. Patent queries are typically small and efficient. See BigQuery pricing for details.

Configuration

The server can be configured using environment variables in your .env file. All settings are optional with sensible defaults:

# API Keys
USPTO_API_KEY=your_key_here

# Google Cloud / BigQuery
GOOGLE_CLOUD_PROJECT=your-project-id
GOOGLE_APPLICATION_CREDENTIALS=/path/to/service-account.json
BIGQUERY_DATASET=patents-public-data:patents
BIGQUERY_LOCATION=US
BIGQUERY_QUERY_TIMEOUT=60
BIGQUERY_MAX_RESULTS=1000

# Logging
LOG_LEVEL=INFO  # Options: DEBUG, INFO, WARNING, ERROR, CRITICAL

# HTTP Settings
REQUEST_TIMEOUT=30.0  # Request timeout in seconds
MAX_RETRIES=3         # Maximum number of retry attempts for failed requests
RETRY_MIN_WAIT=2      # Minimum wait time between retries (seconds)
RETRY_MAX_WAIT=10     # Maximum wait time between retries (seconds)

# Session Management
SESSION_EXPIRY_MINUTES=30  # How long to cache ppubs sessions
ENABLE_CACHING=true        # Enable/disable session caching

# API Endpoints (usually don't need to change)
PPUBS_BASE_URL=https://ppubs.uspto.gov
API_BASE_URL=https://api.uspto.gov

Claude Desktop Configuration

To integrate this MCP server with Claude Desktop:

Update your Claude Desktop configuration file (claude_desktop_config.json):

 {
   "mcpServers": {
     "patents": {
       "command": "uv",
       "args": [
         "--directory",
         "/Users/username/patent_mcp_server",
         "run",
         "patent-mcp-server"
       ]
     }
   }
 }

You can find claude_desktop_config.json on a mac by opening the Claude Desktop app, opening Settings (from the Claude menu or by Command + ' on the keyboard), clicking "Developer" in in the sidebar, and "Edit Config."

Replace /Users/username/patent_mcp_server with the actual path to your patent_mcp_server directory if that's not where it was cloned. (If you're on a mac, this may mean simply replacing username with your username.)

When integrated with Claude Desktop, the server will be automatically started when needed and doesn't need to be run separately. The server uses stdio transport for communication with Claude Desktop or other MCP clients running on the same host.

Available Functions

The server provides the following functions to interact with USPTO data. Note that the Claude Desktop client does not fully support all of these tools. For example, Claude Desktop does not at present allow for download of PDFs.

Public Patent Search (ppubs.uspto.gov)

ppubs_search_patents - Search for granted patents in USPTO Public Search
ppubs_search_applications - Search for published patent applications in USPTO Public Search
ppubs_get_full_document - Get full patent document details by GUID from ppubs.uspto.gov
ppubs_get_patent_by_number - Get a granted patent's full text by number from ppubs.uspto.gov
ppubs_download_patent_pdf - Download a granted patent as PDF from ppubs.uspto.gov (not currently supported by Claude Desktop)

Open Data Portal API (api.uspto.gov)

get_app(app_num) - Get basic patent application data
search_applications(...) - Search for patent applications using query parameters
download_applications(...) - Download patent applications using query parameters
get_app_metadata(app_num) - Get application metadata
get_app_adjustment(app_num) - Get patent term adjustment data
get_app_assignment(app_num) - Get assignment data
get_app_attorney(app_num) - Get attorney/agent information
get_app_continuity(app_num) - Get continuity data
get_app_foreign_priority(app_num) - Get foreign priority claims
get_app_transactions(app_num) - Get transaction history
get_app_documents(app_num) - Get document details
get_app_associated_documents(app_num) - Get associated documents
get_status_codes(...) - Search for status codes
search_datasets(...) - Search bulk dataset products
get_dataset_product(...) - Get a specific product by its identifier

Google Patents Public Datasets (BigQuery)

google_search_patents(query, country, limit) - Search patents by text in title/abstract across 90M+ publications
google_get_patent(publication_number) - Get complete patent details by publication number
google_get_patent_claims(publication_number) - Get all claims for a patent
google_get_patent_description(publication_number) - Get full patent description/specification
google_search_by_inventor(inventor_name, country, limit) - Find patents by inventor name
google_search_by_assignee(assignee_name, country, limit) - Find patents by company/assignee
google_search_by_cpc(cpc_code, country, limit) - Search patents by CPC classification code

Supported Countries: US, EP (European Patent Office), WO (WIPO/PCT), JP (Japan), CN (China), KR (South Korea), GB (Great Britain), DE (Germany), FR (France), CA (Canada), AU (Australia)

Note: Google Patents tools require Google Cloud credentials (see Google Cloud Setup above).

Refer to the function docstrings in the code for detailed parameter information.

Recent Improvements (v0.2.2)

This release includes significant improvements to code quality, reliability, and maintainability:

Architecture & Code Quality

Centralized Configuration - All settings now managed through environment variables with sensible defaults
Constants Module - Magic strings extracted to a dedicated constants module for consistency
Error Handling - Standardized error responses across all endpoints using ApiError utility class
Code Deduplication - Extracted common patent search logic to eliminate ~80 lines of duplicate code
Input Validation - Automatic validation and sanitization of patent/application numbers using Pydantic

Reliability & Performance

Retry Logic - Exponential backoff retry mechanism for network errors using tenacity
Session Caching - ppubs.uspto.gov sessions cached for 30 minutes (configurable) to reduce overhead
Resource Management - Proper cleanup of HTTP clients with context managers and shutdown handlers
Type Hints - Comprehensive type annotations throughout the codebase

Developer Experience

pytest Framework - Modern test framework with async support replacing custom test runner
Python 3.10+ Support - Lowered requirement from 3.13 to 3.10 for broader compatibility
Better Logging - Configurable log levels via environment variables
Development Tools - Added pytest, pytest-asyncio, and pytest-cov to dev dependencies

Testing

The /test/ directory contains test suites for validating the MCP server functionality:

test_tools_pytest.py - Modern pytest-based test suite for all MCP tools (recommended)
test_tools.py - Legacy test runner (still functional)
test_patents.py - Direct HTTP request tests for debugging

Test results in JSON and PDF format are stored in the /test/test_results subdirectory.

Running Tests

# Run all tests with pytest (recommended)
uv run pytest test/test_tools_pytest.py -v

# Run excluding slow tests (like PDF downloads)
uv run pytest test/test_tools_pytest.py -v -m "not slow"

# Run with coverage report
uv run pytest test/test_tools_pytest.py --cov=patent_mcp_server

# Run legacy test suite
uv run test/test_tools.py

Development

To install development dependencies:

uv sync --dev

License

MIT

riemannzeta/patent_mcp_server