mcp-chatgpt-bridge
MCP server that lets Claude Code consult ChatGPT for a second opinion mid-task, with explicit context passed by the caller.
README
mcp-chatgpt-bridge
MCP server that lets Claude Code consult ChatGPT (OpenAI) for a second opinion mid-task, with explicit context passed by the caller.
What it does
When Claude Code is stuck on a design decision, needs a cross-check, or wants an independent opinion, it can call the ask_chatgpt tool to consult OpenAI's ChatGPT. ChatGPT sees only the context you provide — it has no access to your repo, files, or conversation history.
Every consultation is automatically pushed to Telegram/Slack (if configured) so you can review the advice independently later.
Quickstart
See Deployment Guide for full setup instructions.
TL;DR:
npm install && npm run build
export OPENAI_API_KEY=sk-...
claude mcp add chatgpt-bridge -- node /Users/hieuho/Desktop/mcp/dist/index.js
The ask_chatgpt tool
ask_chatgpt({
question: string, // The specific question to ask
context: string, // Self-contained background ChatGPT needs
model?: string // Optional model override (defaults to "gpt-5")
})
ChatGPT responds in Vietnamese with three sections: Câu hỏi (question recap), Khuyến nghị (recommendation), Giải thích (reasoning).
Architecture
Claude Code calls the ask_chatgpt MCP tool via stdio. The server forwards to OpenAI's API (or OpenRouter if you use an sk-or- key). The answer is returned to Claude Code and simultaneously pushed to Telegram/Slack as a side-channel for offline review.
A global Claude Code hook enforces auto-consult: Claude cannot end a turn with a question until it has called this tool first.
See System Architecture for a detailed diagram.
Project Status
- Maturity: Demo / personal use (0.1.0)
- Test suite: None yet
- CI/CD: None yet
- Known gap: No rate-limit or cost-cap enforcement on
ask_chatgptcalls
See Project Roadmap for details.
Documentation
- Project Overview & PDR — Purpose, scope, non-goals
- Codebase Summary — File-by-file breakdown and data flow
- Code Standards — Conventions observed in this codebase
- System Architecture — Architecture diagram and component details
- Deployment Guide — Setup, registration, environment variables
- Project Roadmap — Current state and open items
License
See LICENSE file if present.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。