Remote control your Cursor AI coding sessions via Telegram, Gmail, and Feishu.
Cursor Autopilot sends your full AI chat responses to Telegram (or Email/Feishu), and lets you reply from your phone to continue or redirect the coding session. Zero files are created in your repository — all configuration lives in VS Code settings and the user home directory.
| Step | Action |
|---|---|
| 1 | Open Cursor |
| 2 | Search for Cursor Autopilot in Extensions (Ctrl/Cmd+Shift+X) |
| 3 | Click Install |
| 4 | Open any project folder |
| 5 | Configure your adapter via Settings > search "autopilot" |
git clone https://github.com/j0ntz/cursor-autopilot
cd cursor-autopilot
npm install
npm run compile
npx vsce package
code --install-extension cursor-autopilot-*.vsixRestart Cursor to activate.
All configuration is done through VS Code Settings (no config files in your repo).
Open Settings (Cmd+,) and search for autopilot, or edit settings.json directly:
Each Cursor window can have its own Telegram bot. Set adapter credentials in Workspace settings (Settings > Workspace tab) to override user-level defaults. This means different projects can notify different Telegram chats.
- Create a bot via
@BotFatheron Telegram — save the token. - Get your chat ID from
@userinfobot. - Set
autopilot.telegram.tokenandautopilot.telegram.chatIdin VS Code settings.
See the Setup Guide for Email and Feishu configuration.
Outbound (Cursor → Telegram): At the end of each AI chat turn, the agent pipes its full response directly to telegram-send.py via stdin in a single command — no intermediate files. The script splits the message into 4096-character chunks and sends them to Telegram. Credentials are injected as terminal environment variables ($AUTOPILOT_TG_TOKEN, $AUTOPILOT_TG_CHAT_ID) — each Cursor window gets its own isolated set, so different workspaces can use different bots with no collisions.
Inbound (Telegram → Cursor): The extension polls Telegram for replies and injects them into the Cursor chat composer:
1= Continue2= Stop- Any other text = custom instruction sent directly to the AI
The extension manages these files in your home directory (never in the repo):
| File | Purpose |
|---|---|
~/.cursor/rules/after_each_chat.mdc |
Cursor rule instructing the agent to send responses to Telegram |
~/.cursor-autopilot/telegram-send.py |
Companion script for chunked message sending (reads stdin) |
The extension also sets AUTOPILOT_TG_TOKEN and AUTOPILOT_TG_CHAT_ID environment variables in the integrated terminal for the current window.
Open Command Palette (Cmd+Shift+P) and run Inject text into Cursor Chat to test the injection mechanism.
See the Setup Guide for detailed setup, adapter configuration, and troubleshooting.
Open an issue on GitHub for bugs or feature requests.
We welcome contributions. Please open a GitHub issue to discuss your ideas before submitting pull requests.
MIT — see LICENSE.
{ "autopilot.enabled": true, "autopilot.adapters": ["telegram"], "autopilot.telegram.token": "123456:ABC-DEF...", "autopilot.telegram.chatId": "987654321" }