PatentSafe MCP

What is PatentSafe MCP?

This is a model context protocol server that connects to PatentSafe and retrieves documents.

How to use PatentSafe MCP?

Configure the server in your MCP client (tested with Claude) by specifying the uvx command with the server package from GitHub, the PatentSafe base URL, and your authentication token. The server runs raw Lucene queries against the PatentSafe database.

Key features of PatentSafe MCP

• Connects to PatentSafe to retrieve documents.
• Uses raw Lucene queries against the PatentSafe database.
• Includes a --max-chars flag to limit character output (default 500k).
• Works with any MCP client.
• Installed via uvx from a GitHub repository.

Use cases of PatentSafe MCP

• Querying and retrieving documents from a PatentSafe instance.
• Integrating PatentSafe search results with an LLM for analysis.

FAQ from PatentSafe MCP

What dependencies does PatentSafe MCP require?

It requires Homebrew and uv installed on macOS, plus a PatentSafe instance with an authentication token.

How does PatentSafe MCP handle large document sets?

The server includes a --max-chars flag (default 500k) to limit output, since PatentSafe can return more documents than an LLM can process.

What transport or authentication method does it use?

It uses an authentication token obtained from your PatentSafe user settings; the server is invoked via uvx over a local MCP transport.

Is PatentSafe MCP production-ready?

No, it is under active development and is not production ready.

Where does the query data live?

All queries are executed directly against your PatentSafe instance’s database; no data is stored or cached by the server.