Patents MCP

What tushariitr-19/patents-mcp MCP server does

The tushariitr-19/patents-mcp MCP server gives an MCP-compatible client access to patent records from the Google Patents public dataset in BigQuery. It is aimed at patent discovery and prior-art research, with separate operations for finding candidate patents, inspecting a specific publication, and retrieving claims.

The available tools are:

• search_patents: searches using a keyword, technology area, or inventor name.
• get_patent: retrieves full details for a patent identified by publication number, including CPC codes and citations according to the listing description.
• get_patent_claims: returns the text of a patent’s legal claims.

This separation lets an agent start with a broad search and then request more detail only for selected publications.

How it works

The server is implemented in Go with the official Go MCP SDK. Its entry point starts the MCP service and registers the patent tools. Tool requests are handled through a BigQuery client, which runs queries against the public Google Patents dataset using the configured Google Cloud project.

A typical workflow is to search for a technical concept or inventor, select a publication number from the results, request the complete patent record, and then fetch the claims for closer comparison. Example research prompts in the project cover transformer neural networks, context-aware interfaces, quantum computing patents, and patents associated with Yann LeCun.

The repository separates server setup, tool implementations, BigQuery access, logging, and shared patent models. That structure is relevant if you plan to extend the server with additional patent-related operations.

Setup and configuration

To run tushariitr-19/patents-mcp MCP server, use a Google Cloud project, enable the BigQuery API, and create a service account. The documented roles are BigQuery Job User and BigQuery Data Viewer. Download the service-account JSON key and make it available to the process.

The README describes a source-build installation:

git clone https://github.com/tushariitr-19/patents-mcp
cd patents-mcp
go build -o patents-mcp-server ./cmd/server/

Set these environment variables before starting the resulting binary:

• GOOGLE_APPLICATION_CREDENTIALS: path to the service-account JSON file.
• GCP_PROJECT_ID: Google Cloud project used for BigQuery jobs.
• DEBUG: optional setting for debug logging.

The documented Claude Desktop configuration starts the compiled binary and passes the credential path and project ID through its env object.

Limitations and notes

The data source is the Google Patents public dataset accessed through BigQuery, so use of the server depends on Google Cloud and BigQuery configuration. The README describes BigQuery’s free allowance as 1 TB per month, but actual usage and account conditions should be checked in Google Cloud. No hosted endpoint or prebuilt package is documented; the provided installation path builds the Go executable locally.

The server does not claim to provide legal advice or replace professional patent analysis. Its documented scope is searching patent data, reading patent details, and retrieving claims. The project is MIT licensed. The tushariitr-19/patents-mcp MCP server is therefore best suited to research workflows that can supply Google Cloud credentials and need structured access to this particular patent dataset.