Gemini MCP Server
Claude Code calls Gemini to analyse whole repositories, search with Google Search grounding and read screenshots.
- 141 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- Reads project files and directories and sends them to the Gemini API; the key or service account is kept in the MCP configuration.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add mcp/gemini-mcp-serverDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Reads project files and directories and sends them to the Gemini API; the key or service account is kept in the MCP configuration.
This is third-party code. Review the repository files before installing.
What it does
The MCP server makes Gemini an assistant to Claude Code. Five tools cover search, analysis and ideas: `gemini_search` searches the web with Google Search grounding, `gemini_analyze_codebase` reads a whole project directory in one request, `gemini_analyze_content` reviews code or a document, `gemini_multimodal_query` describes images, `gemini_brainstorm` proposes ideas using the README and PRD. The default model is `gemini-3.1-pro-preview`; quick tasks fit `gemini-3-flash-preview` through the `model` parameter. You can connect with a Google AI Studio key or a Vertex AI service account: the server detects the mode from environment variables.
Who it is for. Claude Code developers who want a second opinion on large code with a long context window and fresh facts from search.
Good fit when
- You need an architecture summary of an unfamiliar repository that does not fit Claude's context.
- A library's documentation came out yesterday and the answer must rely on search.
- A design or diagram is an image and its content must become text.
- You want feature ideas drawn from the project README.
Not a fit when
- The code is under an agreement that forbids sending it to third parties.
- There is no access to the Gemini API.
- Local file search is enough: Claude Code manages without an external model.
Example request
Use Gemini to analyse the `./src` directory: which architectural patterns are used and where the weak spots are.MCP
- Transport
- stdio
- Authentication
- API key
Environment variables
GEMINI_API_KEYrequired, secret- Google AI Studio key. With Vertex AI the service account fields are set instead.
Limitations
You need your own Google AI Studio key or a Vertex AI account with the Vertex AI User role. The Gemini API is unavailable from Russia. Files and queries go to Google, and `gemini_analyze_codebase` sends directory contents in the request. For Vertex AI the service account key fields are pasted straight into the `env` block of the MCP config, where they sit in plain text. Version 2.0.0 was rewritten on the official SDK, so older clients that lost tools now connect normally. The last commit is from July 2026.
How to disable. Remove the `gemini` entry from the MCP configuration and restart Claude Code. Revoke the API key in Google AI Studio or the Google Cloud console.
FAQ
How do I connect Gemini to Claude Code with this server?
Add an MCP entry with `npx -y @lkbaba/mcp-server-gemini` and a `GEMINI_API_KEY` in `env`, then restart Claude Code.
Which model is the default?
`gemini-3.1-pro-preview`. For simple tasks set `gemini-3-flash-preview` in the `model` parameter. The server maps old model names to the current one.
How do AI Studio and Vertex AI differ?
AI Studio works with one key and suits personal use. Vertex AI uses a Google Cloud service account with role-based access; if both modes are configured, Vertex AI wins.
Does the server read my files?
Yes, when you pass a file or directory path to `gemini_analyze_codebase` or `gemini_analyze_content`. The content goes to the Gemini API.
How do I set up Vertex AI?
Copy the service account JSON fields into the `env` block of the MCP config. The server detects `type: service_account` itself, and the account needs the Vertex AI User role.
Discussion
No questions or reviews yet.
Related
MCP serversGoogle's open-source terminal agent on Gemini models: code work, shell, web search and MCP
An MCP server that lets Claude Code, Codex and other agents call Gemini, GPT, Grok and local models for review, debugging and consensus
Plugin and MCP server that asks GPT, Gemini, Grok and local models to review a plan or hunt for bugs before code ships.
Two Claude Code subagents: broad tasks go to Gemini, narrow ones to GPT, and results return to Claude for review.
An MCP server that lets agents in neighboring tmux panes read each other, hand off tasks and watch logs, no framework.
A plugin that gives an agent the IntelliNode skill and MCP server: ask any model, search, review code, tests, SQL, images and speech
README in short
The project grew from a fork of aliargun/mcp-server-gemini. Version 2.0.0 rewrote the protocol layer on the official `@modelcontextprotocol/sdk`, fixing a spec violation that made strict clients drop the connection or miss the tools.