MCP server that exposes OpenAI’s ChatGPT (Responses API) to Claude Desktop, including optional web-search support and conversation state handling.
https://github.com/billster45/mcp-chatgpt-responsesTired of switching between Claude and ChatGPT tabs to compare responses or get different perspectives on your code? This MCP server bridges that gap, letting you call ChatGPT's API directly from Claude Desktop—including web search and proper conversation state management.
You're already using Claude Desktop for development work, but sometimes you need ChatGPT's different approach to problem-solving, or you want access to real-time web search. Instead of context-switching between applications, this server lets you:
Conversation State That Actually Works: Uses OpenAI's Responses API instead of manual history management. Your ChatGPT conversations maintain proper context across multiple exchanges—no more losing thread in complex technical discussions.
AI-to-AI Collaboration: Tell Claude to have a conversation with ChatGPT about your architecture decisions. Watch them debate different approaches while you focus on implementation. It's like having two senior developers discuss your problem.
Real-Time Information: The web search tool means you can ask questions like "What's the current status of the React 19 release?" or "What are today's AWS service outages?" and get accurate, current answers.
Zero Context Switching: Everything happens in Claude Desktop. No more copy-pasting between interfaces or managing multiple conversation threads across different platforms.
Code Review Collaboration: Have Claude analyze your code, then ask ChatGPT for a second opinion on the same implementation. Compare their suggestions for refactoring or optimization approaches.
Research and Planning: Use web search to get current information ("What are the latest Kubernetes security patches?"), then have both AIs discuss implementation strategies based on that fresh data.
Problem-Solving Debates: When stuck on architectural decisions, let Claude and ChatGPT discuss the tradeoffs while you listen in. Their different reasoning patterns often reveal edge cases you hadn't considered.
Learning Complex Topics: Ask ChatGPT to explain a concept, then have Claude build on that explanation or offer alternative perspectives. The back-and-forth often provides deeper understanding than either AI alone.
Setup takes about 5 minutes. Clone the repo, add your OpenAI API key to Claude Desktop's config, restart the app. That's it.
The server runs locally, so your API calls stay on your machine until they hit OpenAI's servers. No third-party service managing your conversations or API keys.
Once configured, you just tell Claude: "Use the ask_chatgpt tool to get ChatGPT's opinion on this database schema." Claude handles the API call and presents the response inline with the conversation.
For extended discussions, the conversation state persists automatically. Tell Claude: "Have a conversation with ChatGPT about microservices vs monoliths," and watch them go back and forth with proper context retention.
The web search integration works the same way: "Use ask_chatgpt_with_web_search to find current TypeScript compiler issues and suggest workarounds."
If you're already using Claude Desktop for development work and occasionally need ChatGPT's perspective or current web information, this eliminates the workflow friction. Instead of managing multiple AI conversations across different interfaces, you get both AI models working together in one place.
The conversation state management alone saves significant time when working through complex problems that require multiple exchanges. And the AI-to-AI discussion feature often surfaces insights neither model would provide individually.
Worth the 5-minute setup if you regularly use both Claude and ChatGPT for development work.