How to Build a Telegram Bot Connected to an AI Agent ?

How to Build a Telegram Bot Connected to an AI Agent (Complete Step-by-Step Guide)

Introduction: The Real Problem That Started It All

After most of my meetings, I needed my AI agents to perform tasks for me:

• Create tickets

• Update my roadmap

• Record decisions

But when a meeting ends, sometimes you don’t have time to open your laptop and write structured notes,

However, I always had time to do one thing:

👉 Open Telegram, record a quick voice message, and explain what needs to happen.

That's why I build this workflow.

This guide will walks you step-by-step through the exact system I built and it is fully open-sourced here:

🔗 GitHub Repository: https://github.com/FlorentLvr/telegram-bot

Prerequisites

Before starting, you’ll need:

• A Telegram account

• Git & GitHub account

• VS Code or Cursor IDE

• ngrok (for HTTPS tunneling locally)

• Render (for deployment)

• OpenAI account (for voice transcription)

• Your AI agent API URL + token

1. Create Your Bot with BotFather

In Telegram, search for @BotFather and send: `/newbot`

You’ll be asked to choose:

1. A name for your bot

2. A username ending with bot (e.g., ai_support_bot)

BotFather returns a Bot Token: 1234567890:ABC-123xyz

Store this securely, it will go in your .env.

2. Clone & Configure the Repository

Clone the repo using command `git clone https://github.com/FlorentLvr/telegram-bot.git cd telegram-bot`

Copy environment template: `cp .env.example .env`

Fill in secrets:

• ENV=dev

• BOT_TOKEN

• OPENAI_API_KEY

• AGENT_API_TOKEN

• AGENT_API_URL

• WEBHOOK_URL

Install dependencies:

pip install -r src/requirements.txt

3. Local Webhook Testing with ngrok

Telegram requires HTTPS — ngrok gives you a quick public tunnel.

Start it and run in your terminal: `ngrok http 10000`

Copy your HTTPS URL (e.g. https://abc123.ngrok.io)
and put it into your WEBHOOK_URL_DEV en your .env.

Run the bot: `python bot.py`

Run the bot: `python telegram_set_webhook.py` to register your webhook in Telegram.

Now send a message to your bot — you should see logs in both:

• Terminal

• ngrok dashboard (http://127.0.0.1:4040)

4. Deploy to Render

1. Create a new Web Service on Render and connect your GitHub repo.

2. Settings:

   • Environment: Python 3

   • Branch: main

   • Root folder: src

   • Build Command: pip install -r requirements.txt

3. Environment variables on Render:

   • BOT_TOKEN = your Telegram token

   • WEBHOOK_URL = https://<your-render-service>.onrender.com (no trailing /webhook)

   • AGENT_API_TOKEN = your AI API token

   • AGENT_API_URL = your AI agent api endpoint

   • OPENAI_API_KEY = your OPENAI_API_KEY

Your bot is now live.

Conclusion

This system was created to solve a practical, daily problem:
How to trigger AI agents immediately after meetings using only your voice.

Now with this bot:

1. Speak your tasks

2. Your bot transcribes them

3. Your agent processes them

4. Tickets instantly appear in your roadmap

It's the fastest workflow I’ve found between thinking → speaking → execution.

Feel free to fork or adapt the repo to your own workflow!