Telegram Notify MCP Server
Connects VS Code Copilot to Telegram for mobile notifications, interactive approval workflows, and remote command input. Enables users to monitor AI agents, approve sensitive operations, and provide follow-up instructions from their smartphone.
README
telegram-copilot-bridge
Control GitHub Copilot CLI remotely from Telegram. Send prompts, manage sessions, and approve tool calls — all from your smartphone.
Architecture
📱 Telegram
↕ (Bot API long-polling)
🐍 telegram-copilot-bridge
↕ (stdin/stdout NDJSON — ACP)
🤖 copilot --acp --stdio
Features
- Remote Prompting — Send Copilot prompts from Telegram, receive results on your phone
- Multi-Session — Run multiple Copilot sessions in parallel, switch between them
- Tool Approval — Approve or deny Copilot's tool calls via inline buttons
- Autopilot Mode — Auto-approve all tool calls for hands-free operation
- Session History — Resume past Copilot sessions with one tap
- Folder Picker —
/newshows inline buttons for project directories
Installation
git clone https://github.com/NobufumiMurata/telegram-copilot-bridge.git
cd telegram-copilot-bridge
pip install -e .
Prerequisites: Install Copilot CLI and authenticate:
winget install GitHub.Copilot # or: npm install -g @github/copilot
copilot # then /login to authenticate
Setup
1. Create a Telegram Bot
- Open Telegram and message @BotFather
- Send
/newbotand follow the prompts - Copy the bot token
2. Get Your Chat ID and User ID
Send any message to your bot, then:
curl "https://api.telegram.org/bot<YOUR_TOKEN>/getUpdates"
From the response, note:
"chat": {"id": ...}→ your Chat ID"from": {"id": ...}→ your User ID (for the allowlist)
3. Configure
Copy the example file and fill in your values:
cp .env.example .env
# .env
TELEGRAM_BOT_TOKEN=1234567890:AAXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
TELEGRAM_CHAT_ID=-100000000000
TELEGRAM_ALLOWED_USERS=123456789
# Default AI model
COPILOT_MODEL=claude-opus-4.6
# Root folder for /dirs and /new folder picker
COPILOT_DIRS_ROOT=/home/user/projects
The .env file is loaded automatically on startup (from the current directory).
Use TELEGRAM_ENV_FILE=/path/to/.env to point to a different location.
Environment variables set in the shell always take priority over .env.
Usage
# Start (reads .env automatically)
python -m telegram_copilot_bridge
# With options
python -m telegram_copilot_bridge \
--cwd /path/to/project \
--model claude-opus-4.6 \
--timeout 120 \
-v
CLI options:
| Flag | Description | Default |
|---|---|---|
--cwd <dir> |
Default working directory for sessions | current dir |
--model <name> |
AI model (e.g. claude-opus-4.6) |
claude-opus-4.6 |
--timeout <min> |
Auto-shutdown in minutes (0 = no timeout) | 0 |
--autopilot |
Auto-approve all tool calls | off (manual approval via Telegram) |
-v, --verbose |
Enable debug logging | off |
Telegram Commands
| Command | Action |
|---|---|
/new [dir] |
Start a new Copilot session (shows folder picker if COPILOT_DIRS_ROOT is set) |
/history [n] |
List past CLI sessions (default: 3) |
/resume <id> |
Resume a past session |
/dirs [dir] |
Browse directories |
/model [name] |
Show/set AI model |
/mode |
Toggle autopilot/manual approval |
/list |
List active sessions |
/switch <id> |
Switch active session |
/status |
Session status |
/stop [id] |
Stop a session |
/done |
Stop all sessions & exit |
/help |
Show commands |
| (any text) | Send as prompt to active session |
Environment Variables
All variables can be set in .env or in the shell. Shell values take priority.
Telegram:
| Variable | Description | Default |
|---|---|---|
TELEGRAM_BOT_TOKEN |
Bot API token | (required) |
TELEGRAM_CHAT_ID |
Target chat ID | (required) |
TELEGRAM_ALLOWED_USERS |
Comma-separated allowed user IDs | (allow all) |
TELEGRAM_ENV_FILE |
Path to the .env file |
.env in CWD |
TELEGRAM_CONFIG_PATH |
JSON credential file (fallback) | — |
Copilot:
| Variable | Description | Default |
|---|---|---|
COPILOT_CLI_PATH |
Path to copilot executable | copilot (from PATH) |
COPILOT_MODEL |
Default AI model | claude-opus-4.6 |
COPILOT_AUTOPILOT |
Auto-approve tool calls (true/false) |
false |
COPILOT_DIRS_ROOT |
Root directory for /dirs and /new folder picker |
(uses --cwd) |
COPILOT_ALLOWED_DIRS |
Comma-separated allowed working dirs | (any) |
COPILOT_ALLOWED_TOOLS |
Comma-separated tools to allow | shell(git),read,write |
COPILOT_PERMISSION_TIMEOUT_SECONDS |
Permission approval timeout (seconds) | 300 (5 min) |
HUB_LOCK_PORT |
TCP port for singleton lock | 47732 |
Security
- User allowlist: Only messages from
TELEGRAM_ALLOWED_USERSare accepted. All other users are silently ignored. - Timeouts: Configurable auto-shutdown timeout (default: no timeout). Permission requests timeout after 5 minutes (configurable via
COPILOT_PERMISSION_TIMEOUT_SECONDS). - No secrets in repo: All credentials via
.envfile or environment variables. - Tool allowlist: Uses
--allow-tool(not--allow-all-tools) to restrict what Copilot CLI can do. - Directory restrictions: Restrict which directories Copilot sessions can operate in via
COPILOT_ALLOWED_DIRS. - Singleton lock: Only one instance can run per machine (TCP port lock).
Limitations
- VS Code ローカル実行は監視不可 — VS Code 上の Copilot が何をしているかを取得する手段はありません。MCP ツールは VS Code → Bridge の一方向呼び出しのみです。
- Bridge 外で起動した CLI のステータスは取得不可 — サーバー上で直接
copilotを起動した場合、リアルタイムの実行状況は取れません。/statusの Activity 表示(events.jsonlの更新時刻による推定)が代替手段です。 - Bridge 外で起動した CLI の承認リクエストは中継不可 — Hub 経由で起動したセッションのみ、ツール承認を Telegram インラインボタンで中継できます。
- シングルインスタンス制限 — Hub は 1 マシンにつき 1 プロセスのみ(TCP ポートロック)。
- Telegram メッセージ長制限 — 4000 文字を超える応答は自動分割されます。
License
MIT
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。
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。
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。
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。
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。
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。
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。
mcp-server-qdrant
这个仓库展示了如何为向量搜索引擎 Qdrant 创建一个 MCP (Managed Control Plane) 服务器的示例。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。