gemini_analyze_code is the one with a purpose beyond 'ask another model': it takes files plus a mode — review, security, testing — so the answer comes back shaped rather than freeform. Run the bundled setup check first, because most failures here are really authentication failures in the underlying tool.
A server that wraps the Gemini command-line tool and exposes prompting, research, code analysis and summarization as four tools.
- gemini_prompt — send a prompt to Gemini and get the response back
- gemini_research — research a topic, optionally with files as context
- gemini_analyze_code — analyze code files for review, explanation, optimization, security or testing
- gemini_summarize — summarize content given as text or as files
- A setup command that checks the command-line tool's installation, authentication and environment before you wire anything in
The Gemini command-line tool installed and signed in on the same machine — the server shells out to it, so an unauthenticated tool means every call fails. Runs as a Python package through uv. This is a community project, not affiliated with Google.
One command — uvx --from git+https://github.com/DiversioTeam/gemini-cli-mcp gemini-mcp