About Reopenai
Reopenai is an MCP server published by jackwener in the AI category.
The repository has 61 stars and 9 forks, with the last commit 7 months ago. Six months or more without a commit doesn't mean the server is broken, but check the open issues before depending on it in production.
Install
git clone https://github.com/jackwener/reopenaiThis server has no ConduID identity, so agent calls to it are not receipted. Pin the version you install and review the source before granting it credentials.
Tools (11)
ai_askSend prompt and track generation state reliably until completion
Parameters not published.
ai_exportBulk export conversation data
Parameters not published.
ai_extract_codeExtract code blocks
Parameters not published.
ai_historyGet extensive conversation history lists
Parameters not published.
ai_list_appsList all supported AI apps and their capabilities
Parameters not published.
ai_modelSwitch active LLM dynamically
Parameters not published.
ai_newStart a new conversation
Parameters not published.
ai_readRead recent conversation history
Parameters not published.
ai_screenshotTake a screenshot of the app's viewport
Parameters not published.
ai_sendSend a message to an AI app (does not wait for resolution)
Parameters not published.
ai_statusCheck CDP connection status
Parameters not published.
Ask AI
Ask AI about Reopenai
Powered by Claude · Grounded in docs
Security checks
- ✓README presentThe repository ships documentation.
5 checks have not been run yet.
README
ReopenAI
Make all AI tools interoperable. Turn any desktop AI client into a standard API endpoint and MCP Server.
ReopenAI acts as a unified bridge, enabling any AI coding assistant (Claude Code, Cursor, Cline, etc.) to freely control and leverage any subscription-based desktop AI app (Antigravity, Cursor, Codex, ChatGPT, ChatWise, Grok).
We offer two distinct integration methods:
Method 1: Native Model Provider (API Proxy) —— 🌟 recommended
This is the most powerful method. It treats your Desktop AI app (e.g., Antigravity) directly as the core backend brain. We launch a local Anthropic-compatible API proxy (/v1/messages) that routes standard inference requests seamlessly into the desktop app via CDP, making Claude Code think it's talking to the official endpoints!
Usage
-
Launch the target AI app (with CDP debugging enabled):
/Applications/Antigravity.app/Contents/MacOS/Electron --remote-debugging-port=9224 -
Start the API Proxy (using opencli):
npm i -g @jackwener/opencli OPENCLI_CDP_ENDPOINT=http://127.0.0.1:9224 opencli antigravity serve --port 8082 -
Connect your tool (Example with Claude Code):
ANTHROPIC_BASE_URL=http://localhost:8082 claude
🎉 Key Proxy Features:
- Auto context isolation by detecting new sessions and actively hitting
New Conversation. - Automatic model translation: transparently maps incoming LLM names (e.g.,
gemini-1.5-pro,claude-3-7-sonnet-20250219) to actual dropdown UI elements and selects them automatically! - Instant generation feedback by natively tracking interface rendering states.
Method 2: MCP Server (Side Tooling)
If you already have a primary foundational LLM, and just want to treat your desktop AI as an external tool/skill (e.g. "Go find X in my ChatGPT history"), you can bridge them using the Model Context Protocol (MCP).
How it Works
Claude Code / Cursor / Cline / ...
↓ MCP Protocol (stdio)
reopenai (this project)
↓ child_process
opencli <app> <command>
↓ CDP Protocol
Electron AI Apps
Quick Start
Add to your .mcp.json:
{
"mcpServers": {
"reopenai": {
"command": "npx",
"args": ["-y", "@jackwener/reopenai"]
}
}
}
Once hooked up, your assistant can natively trigger these types of commands:
- "Check if Antigravity is connected" → calls
ai_status - "Ask Cursor to write a sorting algorithm" → calls
ai_ask - "Read the latest conversation from ChatGPT" → calls
ai_read
Supported Applications
| App | CDP Port | Supported Commands |
|---|---|---|
| Antigravity | 9224 | status, serve(http proxy), send, read, new, extract-code, model, watch |
| Cursor | 9226 | status, send, read, new, ask, model, extract-code, screenshot, history, export |
| Codex | 9222 | status, send, read, new, ask, model, extract-diff, screenshot, history, export |
| ChatGPT | 9224 | status, send, read, new, ask |
| ChatWise | 9228 | status, send, read, new, ask, model, screenshot, history, export |
| Grok | 9234 | ask, debug |
Complete MCP Tools Roster
| Tool | Description |
|---|---|
ai_list_apps |
List all supported AI apps and their capabilities |
ai_status |
Check CDP connection status |
ai_send |
Send a message to an AI app (does not wait for resolution) |
ai_read |
Read recent conversation history |
ai_ask |
Send prompt and track generation state reliably until completion |
ai_new |
Start a new conversation |
ai_model |
Switch active LLM dynamically |
ai_extract_code |
Extract code blocks |
ai_screenshot |
Take a screenshot of the app's viewport |
ai_history |
Get extensive conversation history lists |
ai_export |
Bulk export conversation data |
Development
git clone git@github.com:jackwener/reopenai.git
cd reopenai
npm install
npm run build
License
Apache-2.0
README copy captured when this server was indexed; it may lag the repository. The original is authoritative.