ChatGPT WebSearch MCP

You can run a local MCP stdio server that exposes OpenAI ChatGPT access with web search capabilities, making it easy for Claude Code and other MCP clients to talk to ChatGPT through a standardized interface.

How to use

You connect to the ChatGPT WebSearch MCP by starting the stdio server and then configuring your MCP client to talk to it. Use the CLI flow for Claude Code or add a configuration entry in your client settings to point to the MCP command. Provide your OpenAI API key to authenticate with the OpenAI service.

Option A: Claude Code quick setup. Install and run the MCP through Claude Code so it installs locally in your project scope. This starts the MCP runtime and makes it available for Claude Code to use.

Option B: Configure in settings. Add a server entry that runs the MCP via npx and shares your API key and any preferred defaults. This keeps the server ready for use by any MCP client that supports the standard interface.

How to install

Prerequisites you need before installing and running the MCP server:

• Node.js 22+ is installed on your system
• An OpenAI API key is available to you and will be used to access the ChatGPT models

Install and start the MCP server using the runtime command shown in the usage examples. You can run it directly with your MCP client or add it to your client’s configuration for automatic startup.

# Run the MCP server directly via npx
npx @nekobato/chatgpt-websearch-mcp

Additional configuration and usage notes

Environment variables you can set to customize behavior (placeholders shown):

{
  "mcpServers": {
    "chatgpt_websearch": {
      "command": "npx",
      "args": ["@nekobato/chatgpt-websearch-mcp"],
      "env": {
        "OPENAI_API_KEY": "your-api-key",
        "OPENAI_DEFAULT_MODEL": "gpt-5",
        "REASONING_EFFORT": "medium",
        "SEARCH_CONTEXT_SIZE": "medium",
        "OPENAI_MAX_RETRIES": "3",
        "OPENAI_API_TIMEOUT": "120000"
      }
    }
  }
}

Troubleshooting and tips

If you encounter timeouts or slow responses, adjust the OPENAI_API_TIMEOUT and consider increasing the maxRetries value. For reasoning models, you may enable higher effort to improve answer quality, with corresponding adjustments to timeout behavior.

Available tools

ask_chatgpt

Send a prompt to ChatGPT and receive a response with options to specify model, system prompts, temperature, and control over retries, timeouts, and streaming behavior.