Gemini Bridge

The Gemini Bridge MCP Server acts as a lightweight connector between AI coding assistants and Google’s Gemini AI through the official CLI.

This MCP server enables you to use Gemini’s powerful language model and large context window directly within your existing development workflow without incurring API costs.

Features

• 🔌 Direct Gemini CLI Integration: Connects to Google’s Gemini AI without API costs using the official command-line interface
• ⚡ Simple MCP Tools: Provides two core functions for basic queries and comprehensive file analysis
• 🎯 Stateless Operation: Maintains no sessions, caching, or complex state management for consistent performance
• 🛡️ Production Ready: Includes robust error handling with configurable timeout settings
• 📦 Minimal Dependencies: Requires only mcp>=1.0.0 and the Gemini CLI installation
• 🚀 Easy Deployment: Supports both uvx and traditional pip installation methods
• 🌐 Universal MCP Compatibility: Functions with any MCP-compatible AI coding assistant

Use Cases

• Code Security Reviews: Analyze authentication patterns, security vulnerabilities, and access control implementations across multiple files in your codebase to identify potential security risks before deployment.
• Architecture Analysis: Review database implementations, API designs, and system architecture decisions by comparing different approaches within your project files to make informed technical decisions.
• Legacy Code Understanding: Examine complex or unfamiliar codebases by querying specific functions, analyzing file relationships, and understanding implementation patterns without reading through extensive documentation.
• Multi-File Refactoring Planning: Assess the impact of proposed changes across related files, understand dependencies, and plan refactoring strategies by analyzing how modifications in one area affect other parts of your system.

How to Use It

1. Install and authenticate the Gemini CLI before configuring the MCP server. Run npm install -g @google/gemini-cli to install the command-line interface globally on your system. After installation, execute gemini auth login to authenticate your Google account and verify the setup works by running gemini --version.

2. Install the package directly from PyPI:

ip install gemini-bridge

then add it to Claude Code:

claude mcp add gemini-bridge -s user -- uvx gemini-bridge

3. For development work or custom modifications, clone the repository from GitHub and install it in development mode.

git clone https://github.com/shelakh/gemini-bridge.git
pip install -e .

4. Each MCP-compatible client requires specific configuration syntax. For Claude Code, the installation command automatically handles the configuration.

For Cursor, create or modify the MCP configuration file at ~/.cursor/mcp.json with the server definition including the uvx command and gemini-bridge arguments.

{
  "mcpServers": {
    "gemini-bridge": {
      "command": "uvx",
      "args": ["gemini-bridge"],
      "env": {}
    }
  }
}

VS Code users should create a .vscode/mcp.json file in their workspace with the appropriate server configuration.

{
  "servers": {
    "gemini-bridge": {
      "type": "stdio",
      "command": "uvx",
      "args": ["gemini-bridge"]
    }
  }
}

Other clients like Windsurf, Cline, and Void follow similar JSON-based configuration patterns but may store the configuration in different locations or use slightly different property names.

5. The default 60-second timeout works for most queries, but longer operations require custom timeout values. Set the GEMINI_BRIDGE_TIMEOUT environment variable to specify timeout duration in seconds. For Claude Code, include the environment variable in the installation command. For manual configurations, add the environment variable to the env object in your MCP configuration file.

6. Available Tools:

The consult_gemini tool handles direct queries without file attachments. It accepts a query string, optional directory parameter for context, and optional model selection between “flash” and “pro” variants. Use this tool for general questions about your codebase or conceptual inquiries.

The consult_gemini_with_files tool performs analysis with specific file attachments. It requires a query string, directory path, list of relative file paths, and optional model selection. This tool excels at detailed code reviews, security analysis, and cross-file comparisons where Gemini needs access to actual file contents.

7. Both tools support model selection through the optional model parameter. The “flash” model provides faster responses suitable for general queries and quick analysis tasks. The “pro” model offers more detailed analysis capabilities for complex code review scenarios and architectural decisions.

8. Usage Examples:

For basic code analysis, call consult_gemini with queries like “What authentication patterns are used in this project?” while specifying the appropriate project directory. This approach works well for understanding overall project structure and implementation patterns.

For detailed file reviews, use consult_gemini_with_files to analyze specific files. Pass a descriptive query, the project directory, and an array of file paths to examine. This method excels for security reviews, code quality assessment, and understanding complex implementations across multiple files.

FAQs

Q: What happens if the Gemini CLI is not installed or authenticated?
A: The server will return a clear error message indicating that the CLI is not available or requires authentication. You must install the Gemini CLI using npm and complete the authentication process before the MCP server can function properly.

Q: Can I use this server with multiple AI coding assistants simultaneously?
A: Yes, the server supports universal MCP compatibility and can work with multiple clients simultaneously. Each client connects independently, and the stateless design ensures no conflicts between different client sessions.

Q: What file types and sizes can be analyzed with the file attachment tool?
A: The server can analyze any text-based files that the Gemini CLI supports, including source code, configuration files, documentation, and data files. File size limits depend on Gemini CLI constraints rather than the MCP server itself.