How do I connect OpenClaw to Telegram?

Create a Telegram bot with BotFather to get a bot token, then paste that token into your OpenClaw instance dashboard. Your self-hosted assistant will start responding to messages in that Telegram chat within seconds.

Is the OpenClaw Telegram bot private?

Self-hosting removes the third-party bot vendor: messages flow to your own instance rather than a shared service, and with a local model the content never leaves your server. Be precise about what it does not change, though. Telegram bot messages travel through Telegram and the Bot API is not end-to-end encrypted, so the platform still handles content in transit no matter who runs the bot.

Why does my Telegram bot ignore messages in a group?

Bots are created with Privacy Mode enabled, so in a group they only receive messages that start with a slash command or that reply to or mention the bot. Everything else is never delivered to your server. Turn Group Privacy off in BotFather under Bot Settings, then remove the bot from the group and add it again, because the setting applies when the bot joins.

What happens if my Telegram bot token leaks?

Treat it as a password, because anyone holding it can act as your bot. Run /revoke in BotFather and it is invalidated immediately and replaced with a new one, which you then paste back into your OpenClaw dashboard.

Are there file size limits on a Telegram bot?

Yes. Through the Bot API a bot can download files up to 20 MB and send files up to 50 MB, and a single message caps at 4,096 characters so long answers arrive split. Large videos or long recordings will not reach your assistant, and the failure is quiet rather than loud, so it is worth knowing before you rely on it.

OpenClaw on Telegram: Your Self-Hosted AI Assistant in a Chat

Telegram is the fastest way to put OpenClaw in your pocket. A bot token takes about a minute to create, there is no approval process and no dedicated phone number involved, and once connected you talk to your own self-hosted assistant the same way you talk to anyone else in your contacts.

Why Telegram is the easiest integration

Every messaging platform asks something of you before it will let a program speak on your behalf. WhatsApp wants a phone number and, for anything beyond personal use, a Meta business account. Discord wants an application, an invite URL and a decision about privileged intents. Telegram wants a conversation with a bot called BotFather, and then it hands you a token.

Always in your pocket. Message your assistant like any other Telegram contact, from a phone, a desktop or the web client, with the conversation synchronised across all of them.
No number, no approval. A bot is not tied to a phone number, so it costs you nothing to create and nothing to abandon if you change your mind.
Model-agnostic. The bot is only the surface. Behind it, OpenClaw can be pointed at OpenAI, Anthropic, Google or a local model, and you can change that without touching Telegram at all.
More than text. Voice notes, images and documents all arrive as messages your assistant can act on, which makes a chat window a surprisingly capable interface.

How to connect OpenClaw to Telegram

Create a bot. In Telegram, message @BotFather, run /newbot, and choose a display name and a username ending in bot. BotFather replies with a token that looks like a long string of digits and letters.
Treat the token as a password. Anyone holding it can act as your bot. If it leaks, /revoke in BotFather invalidates it immediately and issues a new one.
Deploy OpenClaw. If you have not already, deploy your instance on Flux in about 30 seconds.
Paste the token. In your OpenClaw dashboard, add the Telegram integration and paste the token.
Start chatting. Open the chat with your bot and send a message. The reply comes from your own instance.

While you are still in BotFather, two optional settings are worth setting now rather than later. /setdescription and /setabouttext control what people see before they start a conversation, and /setcommands gives your bot a command menu, which is the difference between an assistant that feels finished and one that feels like a prototype.

Using it in a group

A Telegram bot behaves differently in a group than it does in a direct message, and the difference catches almost everyone once. Bots are created with Privacy Mode enabled, which means that in a group they only receive messages that start with a slash command, or that reply to or mention the bot directly. Every other message in that group is never delivered to your server at all.

That is why the same bot can answer your direct messages perfectly and appear completely dead in a group. It is not ignoring anyone; it genuinely never heard them. Turn Group Privacy off in BotFather under Bot Settings, then remove the bot from the group and add it again, because the setting takes effect when the bot joins. Our guide on an assistant that stops responding covers this and the equivalent trap on Discord in more detail.

Worth pausing on rather than clicking through, though. Privacy Mode off means the assistant receives everything said in that group. That is exactly what you want for a small household or team chat, and exactly what you do not want in a large one.

What Telegram sees, and what it does not

Self-hosting changes a great deal about privacy, and it is worth being precise about what it changes rather than overselling it.

What it does change: your messages are processed by an instance you control. There is no third-party bot vendor between you and your assistant, no shared service accumulating your conversation history, and no product whose terms can change under you. If you point OpenClaw at a local model, the content of your conversations never leaves your own server at all.

What it does not change: Telegram bot messages travel through Telegram. The Bot API is not end-to-end encrypted, and messages to a bot are not in a Secret Chat, so Telegram itself handles the content in transit regardless of who runs the bot. Self-hosting removes the extra intermediary; it does not remove the platform. For most people that is a good trade and worth making knowingly. If it is not an acceptable trade for a particular conversation, that conversation belongs in the OpenClaw web interface rather than in any chat platform.

Limits worth knowing before you rely on it

The Telegram Bot API has a few boundaries that only become visible when you hit them:

File size. Bots can download files up to 20 MB through the Bot API and send files up to 50 MB. A large video or a long recording will simply not reach your assistant, and the failure is quiet rather than loud.
Rate limits. Telegram limits how fast a bot can send messages, which matters if your assistant produces long streams of output rather than single replies.
Message length. A single Telegram message caps at 4,096 characters, so long answers arrive split across several messages.

None of these is a problem for everyday use. They are the kind of thing worth knowing in advance so that a quiet failure reads as a documented limit rather than as a broken integration.

What people actually use it for

The pattern that makes a chat assistant worth having is not asking it questions, which is what a browser tab already does well. It is the things that benefit from being one message away: forwarding an article and asking for the argument in three lines, sending a photograph of a document and asking what it says, dictating a voice note while walking and getting a written summary back, or triggering something on your own network without opening a laptop. A chat window is a low-friction interface, and low friction is what turns a tool into a habit.

Get started

Telegram is one of 20+ integrations. Explore WhatsApp and Discord too, size your instance with the server requirements, or read the full self-host OpenClaw guide. Ready now? Deploy from the homepage.