Build a Telegram Customer Support Bot in 10 Minutes (No Code)

Why Telegram for support?

Telegram is one of the cheapest, lowest-friction channels you can give a customer. No ticketing app to install, no email to write, no chat widget to load. They open Telegram, message your bot, and get an answer.

Most teams skip it because the setup feels like infrastructure work — bot tokens, webhook servers, RAG plumbing, message queues. With GenMB chat agents none of that exists. You describe the agent, upload your docs, paste a Telegram token, and ship.

What you'll build

By the end of this 10-minute tutorial you'll have:

• A Telegram bot that answers product questions from your docs
• Memory so it remembers each user across conversations
• A "talk to a human" hand-off that emails you the conversation
• A daily summary email of every conversation it had

Total cost on the GenMB Business plan: 1 credit to create the agent, 1 credit per response.

Step 1 — Create the agent (90 seconds)

Open the GenMB workspaces page → "Assistants" tab → New assistant. Pick the "Customer support" preset, or describe what you want in plain English:

"A friendly support agent for Acme that answers product questions, escalates angry users, and remembers what each customer asked before."

GenMB generates the agent's system prompt, personality, and skill stack. You can review the YAML it produces and tweak any field before deploying.

Step 2 — Upload your knowledge base (2 minutes)

Drop your help center articles into the knowledge tab. PDF, TXT, Markdown, CSV — all supported. GenMB chunks them, embeds them with a 768-dim model, and stores them in pgvector for semantic search.

When a user asks "how do I export my data," the agent retrieves the top 3 relevant chunks before answering. The result: answers come straight from your docs, with citations.

Aim for chunks under 1,000 words. If your docs are dense, split them. The retrieval quality scales with chunk specificity.

Step 3 — Connect Telegram (60 seconds)

In Telegram:

1. Message @BotFather → /newbot
2. Pick a name and username
3. Copy the token (123456:ABC-DEF...)

Back in GenMB → "Channels" tab → Telegram → paste the token → Connect. The agent starts polling for messages within seconds.

That's it. You can DM the bot now and get answers from your docs.

Step 4 — Add the hand-off skill

Edit the agent's skills. Toggle on `escalate_to_human` and configure the destination email. Now the agent recognizes phrases like "I want to speak to someone" or sustained negative sentiment and automatically pings you with the conversation transcript.

You don't need to write any logic for this. The skill is built-in.

Step 5 — Schedule a daily summary

Workflows tab → New workflow → "Scheduled". Set cron to 0 9 * * * (9am every day). Add nodes:

1. Database query → SELECT * FROM assistantMessages WHERE created_at > now() - interval '1 day'
2. AI generate → "Summarize the most common topics customers asked about today in 5 bullets."
3. Send email → To: you@yourcompany.com

Activate the workflow. Every morning you wake up to a one-glance brief of yesterday's support volume.

Iteration tips

After a week, look at the conversation logs in the assistant detail panel. The two most useful patterns:

1. Topics the agent declined → either add docs or expand the system prompt to handle them.
2. Repeated questions → these are your top FAQ candidates. Pin them to your help center.

What this replaces

A typical SaaS pays $200–$2,000/mo for Intercom or Zendesk. They get a chat widget on the website. With GenMB you get a real Telegram bot, RAG-grounded answers, hand-off, and daily reporting — for the cost of a Business plan and the credits you actually use.

The difference is that you own the conversation history (it's in your GenMB workspace) and you can extend behavior with workflows that competitors gate behind add-on pricing.

Ship it

Try the Telegram bot template to start from a working agent in one click.