home / mcp / gemini thinking mcp server

Gemini Thinking MCP Server

Provides access to Gemini thinking capabilities for step-by-step reasoning and mathematical problem solving via an MCP client.

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
  "mcpServers": {
    "falahgs-gemini-thinking-mcp": {
      "command": "node",
      "args": [
        "/path/to/gemini-thinking-mcp/build/index.js"
      ],
      "env": {
        "GEMINI_API_KEY": "YOUR_API_KEY"
      }
    }
  }
}

You run a local MCP server that exposes the Gemini reasoning model to your client apps. This server handles requests, talks to the Gemini API, and returns structured step-by-step analyses and answers for tasks like math reasoning and logical deduction.

How to use

To use the Gemini Thinking MCP server, start it locally and connect your MCP client (for example Claude Desktop) to the provided MCP endpoint. In conversations, initiate a request by addressing the Gemini thinking server, for example by typing a message that asks for step-by-step reasoning or a structured solution. The server responds with detailed, reasoned outputs suitable for mathematical problem solving and logical analysis. You can tailor the results by adjusting prompts and the model parameters to influence depth, determinism, and length.

Key usage tips: - Start with a clear, explicit prompt to guide the model’s reasoning (e.g., "Solve this problem step by step and show your reasoning"). - Use max_tokens to cap output length and temperature to control randomness if your client exposes these settings. - Treat the model’s step-by-step explanations as a problem-solving trace, and review each reasoning step for correctness or gaps.

How to install

Follow these concrete steps to install and run the Gemini Thinking MCP server locally.

# Prerequisites
node -v
# Should be v18 or higher
npm -v

# 1) Clone the project
git clone https://github.com/falahgs/gemini-thinking-mcp.git
cd gemini-thinking-mcp

# 2) Install dependencies
npm install

# 3) Create an environment file with your Google Gemini API key
# Replace your-api-key-here with your actual key
GEMINI_API_KEY=your-api-key-here

# 4) Build the project
npm run build

Note: You need a Gemini API key to authorize requests to the Gemini service. Keep your key secure and do not share it publicly.

Additional content

Configuration and security notes: - You will run the MCP server locally and connect it to your client applications. Ensure your API key is kept secure and never embedded in client-side code. - The server exposes an endpoint that your MCP client uses to send prompts and receive reasoning-enabled responses. You can adjust model behavior in your client by setting prompt text and optional parameters such as max_tokens and temperature as supported by your client.