Telegram Bots

ADK-TS Telegram bots use the Model Context Protocol (MCP) to connect your agent to Telegram. The @iqai/mcp-telegram server handles communication with Telegram's Bot API using the Telegraf library. Best for: Chat-based agents, customer support bots, and interactive assistants. This guide covers deploying polling-based Telegram bots to various platforms.

What You'll Need

Before deploying your Telegram bot, ensure you have:

• Telegram Bot Token: Obtained from @BotFather
• LLM API Key: Google AI, OpenAI, Anthropic, or another supported provider
• ADK-TS Telegram Bot Project: Get started with:
  • Use the ADK CLI to create a new project: adk new my-agent --template telegram-bot
  • Or clone the Telegram Personal Assistant sample
• Cloud Platform Account: Railway, Render, Heroku, or Docker hosting

How Telegram Polling Works

Understanding the architecture helps you choose the right deployment settings.

┌─────────────────┐     polls for      ┌─────────────────┐
│   Your Agent    │ ←───────────────── │  Telegram API   │
│  (MCP Server)   │                    │                 │
│                 │ ──────────────────→│                 │
└─────────────────┘   sends responses  └─────────────────┘
        ↑
        │ LLM requests
        ↓
┌─────────────────┐
│   LLM Provider  │
│ (Google, OpenAI)│
└─────────────────┘

Key characteristics of polling mode:

• The MCP server actively polls Telegram's API for new messages
• No incoming HTTP requests — your bot initiates all connections
• No port binding or HTTPS certificates required
• Runs as a long-lived background process

This is why web service deployments fail — they expect your app to respond to HTTP health checks, but polling bots don't listen for incoming requests.

Environment Variables

Configure these environment variables on your deployment platform:

Deploy to Railway

Railway automatically detects long-running processes, making it ideal for Telegram bots.

Step 1: Create Your Project

1. Log in to Railway
2. Click "New" or "New Project"
3. Select "Deploy from GitHub repo" and connect your repository

Step 2: Configure Build Settings

Railway typically auto-detects Node.js projects. Verify or set:

• Build Command: pnpm install && pnpm build
• Start Command: node dist/index.js

Step 3: Add Environment Variables

Go to your service's Variables tab and add the required environment variables:

TELEGRAM_BOT_TOKEN=your_bot_token_here
GOOGLE_API_KEY=your_api_key_here
LLM_MODEL=gemini-2.5-flash
ADK_DEBUG=false  # optional

Step 4: Deploy

Railway deploys automatically when you push to your connected branch. Check the Deployments tab to monitor progress.

Step 5: Verify

Confirm your bot is running correctly:

1. Check the logs for successful initialization
2. Send a message to your bot on Telegram
3. Confirm you receive a response

Deploy to Render

Render requires explicit configuration for background workers.

Step 1: Create a Background Worker

1. Log in to Render
2. Click "Add new" → "Background Worker"
3. Connect your GitHub repository

Step 2: Configure the Service

Configure the following settings for your service:

Step 3: Add Environment Variables

In the Environment section, add the required environment variables:

TELEGRAM_BOT_TOKEN=your_bot_token_here
GOOGLE_API_KEY=your_api_key_here
LLM_MODEL=gemini-2.5-flash
ADK_DEBUG=false  # optional

Step 4: Deploy

Click "Deploy Background Worker". Render will build and deploy your bot.

Step 5: Verify

1. Check the logs for successful initialization
2. Send a message to your bot on Telegram
3. Confirm you receive a response

Deploy to Heroku

Heroku uses a Procfile to define process types. Use worker instead of web.

Step 1: Install Heroku CLI

If you haven't already, install the Heroku CLI:

npm install -g heroku

Then log in to your Heroku account:

heroku login

Step 2: Create Your App

Create a new Heroku app with a unique name:

heroku create your-bot-name

Step 3: Create a Procfile

Create a Procfile in your project root (no file extension) with the following content:

worker: node dist/index.js

Step 4: Add Environment Variables

Set the required environment variables for your bot:

heroku config:set TELEGRAM_BOT_TOKEN=your_bot_token_here
heroku config:set GOOGLE_API_KEY=your_api_key_here
heroku config:set LLM_MODEL=gemini-2.5-flash
heroku config:set ADK_DEBUG=false  # optional

Step 5: Deploy

Push your code to Heroku and scale the worker process:

git push heroku main

Scale the worker process (workers start at 0 by default):

heroku ps:scale worker=1

Step 6: Verify

Check the logs to confirm your bot is running:

heroku logs --tail

Verify the worker process is active:

heroku ps

Deploy with Docker

Docker provides maximum flexibility for deploying anywhere.

Step 1: Create a Dockerfile

Create a Dockerfile in your project root with the following content:

FROM node:20-alpine

# Enable pnpm
RUN corepack enable && corepack prepare pnpm@9.12.0 --activate

WORKDIR /app

# Install dependencies
COPY package.json pnpm-lock.yaml ./
RUN pnpm install --frozen-lockfile

# Build the project
COPY tsconfig.json ./
COPY src ./src
RUN pnpm build

ENV NODE_ENV=production

CMD ["node", "dist/index.js"]

Step 2: Build the Image

Build the Docker image for your target platform.

For local testing, build for your machine's architecture:

docker build -t my-telegram-bot .

For cloud deployment (most platforms use AMD64), specify the platform:

docker buildx build --platform linux/amd64 -t my-telegram-bot .

Step 3: Test Locally

Test your bot locally before deploying.

Run using an env file:

docker run --env-file .env my-telegram-bot

Or pass variables directly:

docker run \
  -e TELEGRAM_BOT_TOKEN=your_token \
  -e GOOGLE_API_KEY=your_key \
  -e LLM_MODEL=gemini-2.5-flash \
  my-telegram-bot

Step 4: Push to a Registry

Tag your image and push it to Docker Hub:

docker tag my-telegram-bot username/my-telegram-bot:latest
docker push username/my-telegram-bot:latest

Step 5: Deploy

Deploy from your registry to Railway, Render, or any Docker-compatible platform by specifying your image URL (e.g., username/my-telegram-bot:latest).

Telegram Troubleshooting

• Service Keeps Restarting: If your bot restarts repeatedly, it's likely deployed as a Web Service instead of a Background Worker. Web services expect a port to be bound, while polling bots do not bind to any port.
• Bot Not Responding: Confirm the TELEGRAM_BOT_TOKEN is valid by contacting @BotFather. Ensure your bot has permission to read messages in group chats if applicable.
• Rate Limiting: Telegram has strict rate limits for bots. Check logs for 429 Too Many Requests errors if your bot is handling high traffic.

Telegram Best Practices

• Separation of Concerns: Keep your bot's logic in the ADK agent and use the Telegram server primarily as a transport layer.
• Error Recovery: Implement a retry strategy for LLM calls to handle intermittent API failures gracefully without crashing the polling loop.

Next Steps