Gemini CLI

2. Gemini CLI Installation and Authentication

This MCP server requires the Gemini CLI to be installed and authenticated:
- Install Gemini CLI: Follow the official installation instructions for your platform
- Authenticate: Rungemini auth loginand follow the prompts
- Verify: Test that it works by runninggemini -p "Hello, world!"

If you see an authentication error, make sure you've completed the login process.

You'll need an MCP-compatible client such as:

- Claude Desktop
- Other MCP-compatible AI assistants

Before running the server, verify your setup:

# Check if everything is configured correctly gemini-mcp setup

- Gemini CLI installation
- Authentication status
- Environment configuration

# Method 1: Run from current directory (for development) uv run gemini-mcp # Method 2: Run using uvx from GitHub uvx --from git+https://github.com/DiversioTeam/gemini-cli-mcp gemini-mcp # Method 3: Run if installed globally gemini-mcp # With debug logging LOG_LEVEL=DEBUG gemini-mcp

# Install from GitHub (use -- separator for proper argument parsing) claude mcp add gemini uvx -- --from git+https://github.com/DiversioTeam/gemini-cli-mcp.git gemini-mcp # Or install from PyPI when published claude mcp add gemini uvx -- gemini-mcp

Add this to~/.config/claude-code/mcp-settings.json:

{ "servers": { "gemini": { "command": "uvx", "args": ](#prerequisites)["--from", "git+https://github.com/DiversioTeam/gemini-cli-mcp.git", "gemini-mcp"] } } }

Add this to your Claude Desktop configuration:

macOS:~/Library/Application Support/Claude/claude_desktop_config.jsonWindows:%APPDATA%\Claude\claude_desktop_config.jsonLinux:~/.config/claude/claude_desktop_config.json

{ "mcpServers": { "gemini": { "command": "uvx", "args": ["--from", "git+https://github.com/DiversioTeam/gemini-cli-mcp.git", "gemini-mcp"], "env": { "LOG_LEVEL": "INFO" } } } }

- prompt(required): The prompt to send
- model(optional): The Gemini model to use (default: gemini-2.5-pro)
- context(optional): Additional context to prepend to the prompt

Research a topic with optional file context.

- topic(required): The research topic or question
- files(optional): List of file paths to include as context
- model(optional): The Gemini model to use

Analyze code files for various purposes.

- files(required): List of code files to analyze
- analysis_type(required): Type of analysis - one of:

- review: Code review for issues and improvements
- explain: Detailed explanation of the code
- optimize: Performance and maintainability suggestions
- security: Security vulnerability analysis
- test: Test case suggestions

- content(optional): Text content to summarize
- files(optional): Files to summarize (alternative to content)
- summary_type(optional): Type of summary - one of:

- brief: 2-3 sentence summary (default)
- detailed: Comprehensive summary
- bullet_points: Bullet point format
- executive: Executive summary for decision makers

Note: Eithercontentorfilesmust be provided.

This MCP server implements several security measures:

- Sandboxed file access: By default, file access is restricted to the current working directory
- Path validation: All file paths are validated to prevent directory traversal attacks
- Configurable allowed directories: You can specify allowed directories via:

- Constructor parameter:GeminiMCPServer(allowed_directories=["/path/to/safe/dir"])
- Environment variable:GEMINI_MCP_ALLOWED_DIRS=/path1:/path2:/path3

- All user inputs are properly sanitized before being passed to the CLI
- Command injection is prevented by using subprocess with argument lists (never shell=True)
- File paths are resolved and validated before use

- Always run the server with minimal required permissions
- Regularly update both the MCP server and Gemini CLI
- Monitor logs for any suspicious activity

git clone https://github.com/DiversioTeam/gemini-cli-mcp.git cd gemini-cli-mcp

You can add the server for local testing with Claude Code CLI:

# Method 1: Add from local directory (for development) claude mcp add gemini-local -- uv run gemini-mcp # Method 2: Add from PyPI (when published) claude mcp add gemini uvx -- gemini-mcp # Method 3: Add from GitHub (development version) claude mcp add gemini uvx -- --from git+https://github.com/DiversioTeam/gemini-cli-mcp.git gemini-mcp # Then you can test the functionality immediately # Example: Send prompts, research topics, analyze code, etc.

# Run all tests with coverage uv run pytest # Run specific test file uv run pytest tests/test_tools.py # Run with verbose output uv run pytest -v

# Run linting uv run ruff check . # Run formatting uv run ruff format . # Run type checking uv run mypy src/ # Run security checks uv run ruff check --select S .

Run the setup check to diagnose common issues:
-

"Gemini CLI not found in PATH"

- Ensure Gemini CLI is installed:which gemini
- Add Gemini to your PATH if needed
- Follow the installation guide for your platform

- The server will detect if you're not authenticated
- Rungemini auth loginto authenticate
- Check if your credentials have expired
- The error message will guide you on what to do

- Check that files are within allowed directories
- Use absolute paths or ensure relative paths are correct
- SetGEMINI_MCP_ALLOWED_DIRSfor custom directories

- Rungemini-mcp setupto check your installation
- Ensure you've installed the package:uv pip install -e .

Enable debug logging to see detailed information:
- Fork the repository
- Create a feature branch (git checkout -b feature/amazing-feature)
- Make your changes and add tests
- Ensure all tests pass and code quality checks succeed
- Commit your changes (git commit -m 'Add amazing feature')
- Push to the branch (git push origin feature/amazing-feature)
- Open a Pull Request

This project is licensed under the MIT License - see theLICENSEfile for details.

- Built on top of theModel Context Protocol
- Integrates with Google's Gemini CLI (unofficial integration)
- Inspired by the MCP ecosystem and community

AI-powered code quality analysis to detect best practice violations, security issues, and architectural problems in real-time.