Telegram Auto-Reply: Claude + Make

A tiny board-game cafe gets the same three questions on Telegram all evening: are you open, can we book a table, do you have a copy of a certain game. The owner is mid-shift refilling water and shuffling decks, so the answers come twenty minutes late, which for a “can we come now” question is the same as no answer.

This build answers the easy ones instantly. Your Telegram bot receives a message, Claude replies within a scope you define, and for anything outside that scope it says a person will follow up rather than guessing. This is the one build in this set that sends on its own, because a chat bot that makes you approve each reply is not a chat bot. The safety lives in the scope, not in a draft step.

We wire it in Make. The shape matches our other guides, a trigger, Claude, an action, but here the trigger and the action are both your bot. The art is in the prompt: a tight scope is what keeps an auto-replying bot from saying something it should not.

What does this workflow actually do?

In one line: an incoming Telegram message gets an instant, scoped reply. Make watches your bot. When someone messages it, Claude answers questions inside the scope you set and defers the rest to a human, and Make sends that reply back in the same chat.

A few real cases, none of them the usual ones:

• The board-game cafe above, answering are-you-open in two seconds instead of twenty minutes.
• A ski-rental shop whose bot handles hours and today’s conditions while the counter is slammed.
• A language-exchange meetup whose bot tells newcomers where and when, in plain terms.
• A maker-space whose bot answers how to get access and whether the laser cutter is free.

The repetitive questions are the ones that interrupt you most. A scoped bot clears those and leaves the real conversations for you.

Why use Make if Claude can already write?

Worth asking, and the answer has a twist here. Claude is good at reading a question and answering it in scope. It is not built to listen to your bot, send replies, and run nonstop. That is plumbing, and Make handles plumbing.

Make does the no-judgment steps: catching each message, sending the reply, staying on. Claude does the one judgment step, the answer. The twist is that this build auto-sends, so there is no human draft to catch a mistake. That puts the whole safety burden on the prompt scope. Make is the wire, Claude is the responder, and your scope is the guardrail. Get the scope right and an auto-reply bot is a help; leave it open and it will confidently answer things it should not.

What do you need before you start?

• A Make account. The free 1,000 operations a month covers a steady stream of messages.
• A Claude API key from the console, not the chat app. See how to use Claude.
• A Telegram bot. Message @BotFather in Telegram, create a bot, and copy the token it gives you.
• Ten minutes to write a tight scope. This is the real work of this build.

Two Make words. A scenario is the whole listen-and-reply automation. A module is one box. Three boxes, and two of them are your bot.

How does the workflow work, step by step?

Three modules, left to right:

Module	App	What it does
1. Trigger	Telegram	Watches your bot and fires on each incoming message
2. Brain	Claude	Answers in scope, defers anything else to a human
3. Output	Telegram	Sends the reply back in the same chat

Step 1: Watch your Telegram bot

Create a scenario and add Telegram Bot, Watch Updates. Connect it with the token from BotFather. This trigger fires whenever someone messages your bot and hands the next step the message text and the chat ID, which is how the reply gets back to the right person.

Step 2: Let Claude answer in scope

Add a Claude module, Create a Prompt, paste your key, pick Sonnet. The scope is the whole game:

You are the assistant for The Meeple Cafe. Reply to this Telegram
message in one or two friendly sentences.
Message: {{text from Step 1}}

Answer ONLY questions about opening hours, booking a table, and which
games we stock. For anything else, say a team member will follow up
shortly and do not guess. Never invent prices, dates, or stock.

Name the bot, list exactly what it may answer, and tell it to defer everything else. A narrow scope is what makes an auto-replying bot safe.

Step 3: Send the reply

Add Telegram Bot, Send a Reply Message, and map the chat ID from Step 1 plus Claude’s text. Mapping the chat ID from the incoming message, not a fixed one, is what sends the reply to the right person. Also filter Step 1 so the bot never reacts to its own messages, or it can talk to itself in a loop.

How do you run it and check the result?

Click Run once, then message your bot from your own Telegram. When I tested this with a throwaway bot, replies came back in a couple of seconds. The tuning that mattered was tightening the scope: my first version cheerfully tried to answer questions it had no business touching, and adding the hard “answer ONLY” line fixed it. Once it stays in its lane, switch the scenario on so it listens around the clock.

After that, the easy questions answer themselves the moment they arrive, day or night, and the messages that actually need you are the only ones left waiting.

What does this cost to run?

Piece	Free tier	If you outgrow it
Make	1,000 operations/month free	Core plan from about $9/month
Claude API	Pay per use	A one-line reply on Sonnet costs a fraction of a cent
Telegram	Free	Free, including the bot
BotFather	Free	Free

Each message is three Make operations and a sliver of Claude. A busy bot can use operations quickly, so watch your volume on the free plan. See tiers in our Make guide.

What can go wrong, and how do you avoid it?

• It answers things outside its scope. Tighten the prompt with a hard “answer ONLY” list and a defer-to-human line. This is the one I had to fix.
• Replies go to the wrong person. Map the chat ID from Step 1, never a fixed chat.
• The bot talks to itself. Filter the trigger so it ignores its own messages, or you get a loop.
• The bot seems dead. Check the BotFather token and that someone has pressed Start on the bot.

The same trigger-brain-action shape runs many builds. See Make AI scenarios for more.

How do you build this in Zapier or n8n instead?

Same three jobs, different names.

Job	Make	Zapier	n8n
Catch a message	Watch Updates	New Message trigger	Telegram Trigger node
Answer in scope	Claude module	Claude (Anthropic) action	Anthropic node
Send the reply	Send a Reply Message	Send Message action	Telegram node

Make and Zapier are easiest to start with. n8n suits self-hosting. Compare them in Zapier vs Make vs n8n.

Common questions about this build

Does this one really send without me?

Yes, this is the auto-reply build. That is why the prompt scope matters so much: it is the only guardrail, so keep it narrow.

How do I make a Telegram bot?

Message @BotFather inside Telegram, send /newbot, follow the prompts, and copy the token. It takes about two minutes.

What stops it from saying something wrong?

A tight scope. You tell Claude exactly what it may answer and to defer everything else. Test it hard before going live.

Can it hand off to a person?

Yes. The defer line tells the user a human will follow up. You can also post those messages to a channel so someone sees them.

Is the chat app the same as the API?

Same models, different door. Make talks to the API, so you need a key from the console, not the chat login.

Sources and references

• Creating a scenario (Make Help)
• Claude API overview (Anthropic)
• Telegram Bots: an introduction (Telegram)
• Claude on Make (integrations)
• Telegram — Grokipedia

Last reviewed: May 2026. Make, Claude, and Telegram change their interfaces; check the official pages for exact button names.

You may also like

• Draft WordPress posts from a spreadsheet with Claude + Make
• Turn a YouTube video into a WordPress draft with Claude
• Turn a blog post into LinkedIn and X posts with Claude
• Make.com Complete Beginner’s Guide
• How to use Claude AI
• Make AI Scenarios: automations that save hours
• The AI Tools Directory