Telegram Bots: From Notification Channel to Internal Assistant

A Telegram bot is the first contact with n8n for many teams: a simple automation sends a message as soon as a form comes in, a server goes down, or an order is completed. That is exactly where this guide starts, because the path from a silent notification channel to a responsive digital assistant is shorter than it looks. n8n provides two dedicated Telegram nodes for this: a trigger that reacts to incoming messages, and an app node that sends messages, files, and chat actions. Combined with an AI agent, the pure sending channel becomes an assistant that answers questions, suggests actions, and if needed waits for approval before executing anything.

As of July 2026, NordFlux regularly uses this pattern in client projects, because it requires no additional server costs and feels familiar to you as a team member: communication runs in an app that is already installed on every smartphone anyway.

The simple start: Telegram as a notification channel

Before n8n can even send a message, it needs a bot account with Telegram itself. You create this via the official BotFather, according to the n8n documentation on Telegram credentials in a few steps:

• Start a chat with the BotFather and enter the command `/newbot`
• Assign a display name, which can still be changed later
• Set a username that must be between 5 and 32 characters long, may only consist of Latin letters, numbers, and underscores, and must end in "bot"
• Copy the generated access token and store it in n8n as a Telegram credential

You then enter this token in both Telegram nodes, in the trigger as well as in the app node. This establishes the connection between n8n and your bot, without you having to operate your own infrastructure.

From trigger to action: what the Telegram nodes can do

The Telegram Trigger starts a workflow as soon as a matching update arrives. According to the documentation, it covers more than twenty event types, ranging from simple text messages and callback queries from inline keyboards to chat joins and reactions to messages. By default, it reacts to all updates except changes to chat members and message reactions, because according to n8n these generate a particularly high number of update calls. In addition, the trigger can be filtered by chat and user IDs, so that only certain people or groups are allowed to trigger a workflow at all.

The associated Telegram-Node then takes over the sending. It supports message, chat, file, and callback operations: text, photos, videos, documents, stickers, location data, and entire media groups can be sent, and messages can be edited, pinned, or deleted afterwards. According to the documentation on the message operations, text is subject to a limit of 4096 characters after formatting, captions are limited to 1024 characters, and Telegram itself allows a maximum of 30 messages per second per bot. For most message operations, the bot must also have been actively added to the respective channel or group, otherwise sending will fail.

The next stage: Telegram as an interface to an internal AI assistant

Once the trigger is running reliably, the next step is obvious: instead of sending a fixed message, you pass the content of an incoming Telegram message to an AI agent in n8n. The agent reads the question, accesses connected tools such as a database, a calendar, or an internal system if needed, and sends the answer back into the chat via the same Telegram node. For your team, this feels like a conversation with a colleague, while technically it is a workflow made up of a trigger, agent, and response node.

This pattern becomes especially practical when the AI assistant not only answers but is also meant to carry out actions, such as rescheduling an appointment or triggering an order. For exactly such cases, n8n offers a built-in approval feature that keeps the assistant on a short leash.

Approvals directly in the chat: Human-in-the-Loop via Telegram

So that an AI agent does not carry out actions without control, Telegram can be integrated directly as an approval channel according to the n8n documentation on Human-in-the-Loop for tools. In the tools section of an agent node, you select the human review step for this, link it to Telegram, and define which tools require confirmation before they run. A human reviewer then sees in Telegram which tool the agent wants to use and with which parameters, and can confirm or reject the action. On confirmation, n8n executes the tool with the proposed values; on rejection, the action is cancelled and nothing further happens in the background.

This way you retain control over critical steps, without having to handle every routine task manually. The digital assistant does the groundwork, you decide on the final step, directly from the Telegram app.

Limitations and typical pitfalls

• One bot token per environment: Anyone who runs the test and production systems with the same bot token risks test messages ending up in the real channel. A dedicated bot per environment creates a clear separation here.
• Plan for rate limits: Telegram limits sending to around 30 messages per second per bot. For mass sending to many chats, a small delay in the workflow is worthwhile.
• Check bot permissions: For most message operations, the bot must be actively invited to the channel or group; for administrative actions such as setting a chat title, it additionally needs admin rights.
• Do not forget filtering: Without a chat or user ID filter in the trigger, the workflow reacts to every message the bot receives. In open groups, this can quickly lead to unwanted triggers.

If you would rather not implement this setup alone: NordFlux takes care of the setup, security, and extension of such Telegram workflows as part of n8n automation for small and medium-sized businesses.

Frequently asked questions

Do I need programming skills for a Telegram bot in n8n?

No. Triggers, actions, and the connection to the AI agent are configured in the n8n editor using form fields and connections between nodes. For more complex logic, some experience with expressions within n8n helps, but you do not have to write your own bot program.

How do I get the access token for my bot?

You get the token directly from Telegram's own BotFather, by creating a new bot there with the command `/newbot`. The BotFather then automatically generates a token, which you enter unchanged into the Telegram credential in n8n.

Can the same bot both send notifications and answer questions?

Yes. A single bot can send messages via the Telegram node and, at the same time, receive incoming messages via the Telegram Trigger. Whether this results in a pure sending channel or a full assistant depends solely on whether an AI agent is connected in between in the workflow.

How secure is an approval of actions via Telegram?

The approval runs via a chat with a fixed chat ID or a specific channel. Only those with access to this chat see the request and can confirm or reject it. For sensitive processes, a separate, tightly restricted approval chat is recommended instead of an open group chat.

What does getting started with a Telegram bot in n8n cost?

Creating a Telegram bot via the BotFather is free, as is using the Telegram nodes in n8n. Costs arise at most for the n8n instance itself, for example for hosting or a cloud license, as well as for a connected AI provider if the bot is expanded into a full assistant.