| name | telegram-bridge |
| description | This skill should be used when the user mentions "telegram", "tgbot", "bot", "mobile control", "remote access", asks to start/stop the bot, wants to send a Telegram notification, or needs to interact with Claude from their phone. Also activates when asking about voice transcription, Whisper, or Turkish voice commands. |
| version | 1.1.0 |
Telegram Bridge Skill
Bidirectional Telegram bot integration for Claude Code.
Capabilities
- Text commands: Send any task to Claude via Telegram
- Photos: Send images for analysis
- Documents: Send text/code files for processing
- Voice: TR/EN voice messages transcribed via Whisper
- Inline keyboard: Quick access buttons (Status, Projects, Log, Stop)
Bot Commands
| Command | Description |
|---|
/run <task> | Execute a Claude task |
/status | Show current project and time |
/projects | List ~/Projects/ directories |
/cd <name> | Switch active project |
/log | Show last 30 log lines |
/stop | Stop the bot |
| Free text | Directly sent to Claude as task |
Setup Requirements
Add to ~/.claude/secrets/secrets.env:
TELEGRAM_BOT_TOKEN=<your-bot-token>
TELEGRAM_CHAT_ID=<your-chat-id>
Get token from @BotFather, chat ID from @userinfobot.
Starting the Bot
/tgbot start
nohup bash ~/.claude/config/telegram-poll.sh ${PWD} > ~/.watchdog/telegram.log 2>&1 &
Architecture
Telegram → getUpdates polling → tg_parse.py → bash handler → claude -p → tg_send.py → Telegram
telegram-poll.sh: Main loop, handles all message types
tg_parse.py: Parses Telegram API responses, handles UTF-8/emoji
tg_send.py: Sends messages with optional inline keyboards
tg_voice.py: Whisper transcription (openai-whisper → whisper.cpp → OpenAI API)
When Absent
When the user is away and Claude needs input, use telegram-ask.sh to wait for a Telegram reply before proceeding.