Pi Telegram Bot

A Telegram bot powered by the pi coding agent — any AI model (Claude, GPT, Gemini) with tools, streaming, and persistent sessions per chat.

Features

• 🤖 Full pi agent (Claude) with all installed tools active
• 💬 One persistent session per Telegram chat (survives restarts)
• ⚡ Live streaming — bot message updates as Claude thinks
• 🗜️ /compact to compress long conversations
• 🔄 /new to start fresh
• 🛑 /abort to cancel a running task
• 🔒 Optional allowlist via ALLOWED_CHAT_IDS

Setup

1. Get a Telegram Bot Token

1. Message @BotFather on Telegram
2. Send /newbot and follow the prompts
3. Copy the token

2. Configure .env

cp .env.example .env

Edit .env:

TELEGRAM_BOT_TOKEN=your_bot_token_here
ANTHROPIC_API_KEY=sk-ant-your-key-here

# Optional: restrict to specific chat IDs
# Get yours by messaging @userinfobot on Telegram
ALLOWED_CHAT_IDS=123456789

3. Install dependencies

npm install

4. Run

# Development (auto-restart on change)
npm run dev

# Production
npm start

Commands

Command	Description
/start	Introduction and help
/new	Start a fresh conversation
/compact	Compress the conversation history
/abort	Cancel the current task
/status	Show session info (model, message count)

Sessions

Each Telegram chat gets its own persistent pi session stored in:

~/pi-telegram-bot/sessions/<chat_id>/

Sessions survive bot restarts and pick up where they left off.

Installed Pi Tools (from nicobailon)

All tools are loaded from ~/.pi/agent/settings.json and activate automatically:

Package	What it does
pi-interactive-shell	Control interactive CLIs in an overlay
pi-design-deck	Visual design deck with multi-slide previews
pi-mcp-adapter	Token-efficient MCP adapter
pi-interview	Interactive forms and user input
pi-web-access	Web search and content extraction
pi-subagents	Async subagent delegation
pi-messenger	Multi-agent communication
pi-review-loop	Automated code review loop
pi-annotate	Visual browser feedback
pi-model-switch	Agent-driven model switching
pi-rewind-hook	Rewind file changes
pi-powerline-footer	Powerline status bar
pi-prompt-template-model	Model frontmatter in prompt templates
pi-skill-palette	VS Code-style skill selector
surf-cli	Agent-controlled Chrome browser
visual-explainer	Rich HTML visual explanations
pi-foreground-chains	Multi-agent workflow orchestration

Deployment (VPS / Fly.io)

With pm2:

npm install -g pm2
pm2 start "npm start" --name pi-bot
pm2 save

With Docker:

FROM node:22-alpine
WORKDIR /app
COPY package*.json .
RUN npm ci
COPY . .
CMD ["npm", "start"]

docker build -t pi-telegram-bot .
docker run -d --env-file .env pi-telegram-bot