Token validated on the spot
Paste the token and the platform calls Telegram's getMe right away. A bad token comes back as an inline form error - never a poller that silently never talks to anyone.
How to build an AI Telegram bot with no code
Describe what the bot should do in a prompt, paste a token from @BotFather into the agent's form in the Dada Cloud console, and within a couple of minutes your own AI bot is answering in Telegram. No lines of code, no polling script to host, no webhook or separate app for the wiring.
Step by step: from a prompt to a bot in Telegram
Four steps, zero lines of code.
Create an agent
In the console's Agents section, describe the bot's role in the system prompt and pick MCP tools if you need them. Save the agent - the Telegram bot field only appears once it exists.
Get a token from @BotFather
Open a chat with @BotFather in Telegram, create a bot with /newbot, and copy the token it gives you.
Paste the token and connect
In the agent's form, paste the token into the Telegram bot field and click Connect. The platform checks it against Telegram's getMe right away and shows an inline error if it's wrong.
Done, message the bot
Once the form shows Connected as @your_bot, message it on Telegram: your message goes to the agent and the model's reply comes back in the same chat.
Why you don't need to write any code
Long polling, not a webhook
The gateway polls Telegram's getUpdates itself. No domain, no TLS certificate, no extra web server to run just for the bot - there is nothing separate to deploy.
One click, no extra app
Binding a bot happens in the same form where you already create and edit the agent - no new console page, no second deployment for the integration.
Disconnect is always safe
The disconnect action works even when there is no existing binding. Calling it twice does nothing the second time, instead of erroring out.
You hear about it if the agent isn't ready
If a message arrives while the agent is still deploying, the gateway retries with backoff and after about 30 seconds of failures tells the user in the chat instead of staying silent.
Gateway restarts don't lose bindings
Every binding lives in the database. A gateway pod restart rebuilds every poller on boot by itself - no manual re-bind needed.
How to build an AI Telegram bot - FAQ
Do I need to know how to code?
No. You need a text prompt describing the agent's role and a token from @BotFather - both go into a plain form in the console, no programming required.
Is this a webhook or polling?
Polling - the gateway calls Telegram's getUpdates itself. No domain, TLS certificate, or separate HTTP server is needed for the bot.
What happens if the agent isn't ready when a message arrives?
The gateway retries with backoff. If the agent isn't up after about 30 seconds, the Telegram chat gets a message saying the agent is still deploying, instead of silence.
What if I paste the wrong token?
The form validates the token against Telegram's getMe before saving anything. A bad token returns an inline error immediately and no binding is created.
Can one bot answer for multiple agents?
No, the binding is strictly one to one: one bot token, one agent. For a second bot, create another one with @BotFather and bind it separately.
Can one agent answer in multiple bots?
Also no - the binding is one to one in both directions. A second bot needs either a separate agent or disconnecting the first binding before making a new one.
What happens to the binding when the gateway restarts?
Nothing is lost - the gateway stores bindings in its database and rebuilds every poller on startup. No manual re-bind needed.
I lost my bot token, what now?
Disconnect the current binding, get a new token from @BotFather (you can revoke the old one), and bind again - the same steps as the first time.
Do I need to write my own polling or webhook code?
No. Normally wiring a Telegram bot to an LLM means writing and hosting your own process, something like aiogram, and keeping it running. Here that part is a built-in feature for agents you already run on the platform.
Give your code a public address
Create an account and deploy your first app. The Free plan costs $0.