Deploy a Telegram AI assistant in 60 seconds

Most AI chatbot tutorials start with 400 lines of boilerplate, a Docker compose file, and a prayer. OpenClaw skips all of that. You pick a model, pick a channel, and deploy. The infrastructure is handled.

#What you need

• A Google account for sign-in
• A Telegram account (to test your bot)
• Optionally, your own API key from OpenAI, Anthropic, or Google

#The flow

Open openclaw.com, sign in with Google, and you land on the deploy screen. Three choices: model, channel, and API key mode. Select Claude Opus 4.5, Telegram, and "We provide" if you don't have your own key yet.

Hit deploy. OpenClaw provisions a secure server, wires up the Telegram Bot API webhook, and hands you a link. Open Telegram, find your bot, send a message. It responds.

#What happens under the hood

OpenClaw spins up an isolated container on managed infrastructure. Your bot gets its own process, its own memory, and its own rate limits. The API key is injected as an environment variable on the server — it never reaches the client device or Telegram's servers.

Incoming messages from Telegram hit the webhook endpoint, get routed to the AI model, and the response is sent back through the Bot API. Latency is typically under 2 seconds for the first token.

#Switching models later

You can swap the underlying model at any time from the dashboard. Switch from Claude to GPT-5.2 or Gemini 3 Flash without redeploying. The bot stays live, the Telegram handle stays the same, and your conversation history is preserved.

That's it. No YAML, no CI pipeline, no infrastructure to babysit. Deploy and move on.