Home / MCP / ChatGPT MCP Server

ChatGPT MCP Server

Provides a local ChatGPT-powered MCP server exposed via stdio for client integration and debugging.

python

14stars

View on GitHub

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
    "mcpServers": {
        "chatgpt": {
            "command": "uvx",
            "args": [
                "mcp-server-chatgpt-app"
            ]
        }
    }
}

You can run and connect to the ChatGPT MCP server locally, then use MCP clients to interact with it. This server exposes a local, stdio-based MCP endpoint that you launch with a simple command, letting you manage ChatGPT-powered capabilities from your development environment or editor integrations.

How to use

Set up and connect to the local ChatGPT MCP server from your MCP client. You will run the server locally and configure your client to communicate over stdio. Follow these patterns to enable smooth operation, keep the server under control, and avoid multiple concurrent instances.

Starting and configuring from your MCP client

Configure the local MCP server as a stdio transport so your client can start and stop it directly. Use the following two options to integrate, depending on the tool you are using.

MCP client configuration (stdio transport)

{
  "type": "stdio",
  "name": "chatgpt",
  "command": "uvx",
  "args": ["mcp-server-chatgpt-app"],
  "env": {},
  "disabled": false,
  "autoApprove": []
}

MCP Inspector and editor tooling integration

If your editor provides an MCP Inspector or a dedicated inspector tool, you can point it to the local server using stdio with the uv transport and the server script. Use this configuration to run the inspector from your editor.

{
  "type": "stdio",
  "name": "mcp_chatgpt_inspector",
  "command": "uv",
  "args": ["--directory", "/Users/<your-username>/Developer/mcp-server-chatgpt-app/src/mcp_server_chatgpt", "run", "server.py"],
  "env": {}
}

Notes on running multiple servers

Be aware that certain clients may not automatically close a running MCP server when closed themselves, potentially leaving multiple MCP servers active. If you suspect this is happening, manually terminate all processes related to the server with a task-kill command such as pkill -f 'mcp-server-chatgpt-app' and then restart the server as needed.

Available tools

MCP Inspector

Provides a debugging interface for the MCP server over stdio by running the inspector against the local server directory.

Chatwise integration

Facilitates adding and configuring MCP server connections in your editor with stdio transport and explicit caution about multiple running instances.