Telegram Bot Builder

Telegram's got over 1B monthly active users. LLMs make it possible to build bots that handle customer messaging for companies. This platform lets you build those bots without writing code.

Comment

Inspiration

At one of my startups, we spent time and money building a mobile app where users could send voice messages that turned into posts on a shared blog. It flopped. Later I realized we could’ve tested the idea way cheaper—just by setting up a WhatsApp or Telegram number and having a person handle messages manually. No app needed.

Now with LLMs, we don't even need a human to run the test. Just hook an LLM into the Telegram API, give it some instructions, wire up a few tools, and let it handle the chats. You just monitor how the conversations go.

What it does

The platform lets you build a Telegram bot by submitting a bot API token from BotFather and plain English instructions for the bot’s behavior. That’s all you need to get rolling.

How it is built

I used Supabase for the backend (separate post on why it's one of the better serverless picks for AI streaming: https://medium.com/@mitek99/build-an-ai-powered-chatbot-with-vercel-ai-sdk-and-supabase-edge-functions-4cd051eac9ec).

Telegram webhooks and messaging run through an edge function built with Claude Code, inside Windsurf and Cursor. I stick to IDEs for backend work.

Frontend’s built with Bolt.

Challenges I ran into

Biggest issue: no time. This is a side project, and I’ve got a full-time job, a family, and just picked up a gravel bike.

Accomplishments that I'm proud of

What's next for Telegram Bot Builder

  • Dig deeper into the Telegram API and handle more use cases, like bots in group chats
  • Port a few popular MCP servers to Webtools and get them running in the cloud
  • Choose a relevant domain name

Built With

  • bolt
  • deno
  • supabase
Share this project:

Updates