Model Context Protocol moderate risk

Chatgpt Responses

This MCP server provides access to OpenAI's ChatGPT API, enabling conversational AI capabilities within Claude Desktop with web search and conversation state management.

Connections & Capabilities

Connects To

GitHubFilesystem

Capabilities

readwriteexec

Quickstart

Install

npx -y @smithery/cli install @billster45/mcp-chatgpt-responses --client claude

Config

{
  "mcpServers": {
    "chatgpt": {
      "command": "uv",
      "args": [
        "--directory",
        "\\path\\to\\mcp-chatgpt-responses",
        "run",
        "chatgpt_server.py"
      ],
      "env": {
        "OPENAI_API_KEY": "your-api-key-here",
        "DEFAULT_MODEL": "gpt-4o",
        "DEFAULT_TEMPERATURE": "0.7",
        "MAX_TOKENS": "1000"
      }
    }
  }
}

Exposed MCP Tools (2)

moderate

ask_chatgpt(prompt, model, temperature, max_output_tokens, response_id)

Sends a prompt to ChatGPT and returns the response.

Allows arbitrary prompt execution, which could lead to unintended actions or information disclosure.

high

ask_chatgpt_with_web_search(prompt, model, temperature, max_output_tokens, response_id)

Sends a prompt to ChatGPT with web search enabled and returns the response.

Combines arbitrary prompt execution with external web access, increasing the risk of accessing malicious content.

Safety Assessment

This MCP server is relatively safe for general use, but the risk of compromised API keys and the potential for misuse through the `ask_chatgpt_with_web_search` tool should be considered. It is safe when used with appropriate rate limiting and monitoring of API usage. It is risky if the OpenAI API key is exposed or if the web search tool is used without proper safeguards.

Uses OpenAI's Responses API for conversation state management, reducing local data handling.
Relies on OpenAI's security measures for API access.
Configurable model versions and parameters limit potential misuse.
Requires an OpenAI API key, providing some level of access control.

Requires an OpenAI API key, which, if compromised, could lead to unauthorized usage.
Exposes the ChatGPT API to Claude Desktop, potentially allowing for unintended actions if not properly controlled.
The `ask_chatgpt_with_web_search` tool introduces the risk of accessing and processing untrusted web content.
Lack of built-in rate limiting could lead to excessive API usage and unexpected costs.