Open-source scaffold CLI for running a Telegram bot that forwards messages to Gemini CLI.

• Python Telegram runtime (bot.py) with:
  • retries + timeout handling
  • per-chat session pinning
  • /sessions, /sessionuse, /sessionclear, /sessiondelete, /continue
  • /cli command-chain mode for fast Gemini CLI commands
  • Telegram chat-mode guardrails for normal messages
• .env runtime configuration
• user systemd service template

• Linux with systemd --user (recommended)
• Node.js >= 20
• Python >= 3.10
• Gemini CLI installed and authenticated (gemini --list-sessions works)
• Telegram bot token from BotFather

npx create-gemini-telegram-agent init \
  --dir ./my-gemini-bot \
  --bot-token-file ~/.config/gemini-telegram-agent/telegram-bot-token.txt \
  --chat-id <your_chat_id>

Alternative token input:

• --bot-token <token> (quick setup; appears in shell history)
• TELEGRAM_BOT_TOKEN=<token> ... init (recommended for CI/non-interactive)

Then verify:

cd ./my-gemini-bot
npx create-gemini-telegram-agent doctor --dir .

create-gemini-telegram-agent init [options]
create-gemini-telegram-agent doctor [--dir <path>]
create-gemini-telegram-agent start [--dir <path>]
create-gemini-telegram-agent sync-commands [--service-name <name>]
create-gemini-telegram-agent uninstall [--service-name <name>] [--purge] [--yes]

• Bot token written to ~/.config/gemini-telegram-agent/telegram-bot-token.txt with 0600
• No token values printed in CLI logs
• Normal Telegram messages run non-YOLO mode by default

npm run check
npm run pack:dry-run
npm publish --access public

• A Telegram bot username/ID alone is not enough for setup. Users need the bot token.
• For OAuth-backed Gemini usage, authentication happens on each host machine.