There is a category of ClawHub skill that does not make headlines but quietly becomes one of the most-used things on a user’s machine. The Telegram skill is the clearest example: it has no flashy AI feature, no self-improvement loop, no 860-integration catalog. What it does is straightforward — it connects your OpenClaw agent to Telegram so you can message it from anywhere and receive replies on your phone. That simplicity is apparently what the market wants. With approximately 145,000 installs, it is the second most-installed skill in the entire ClawHub registry.
The Problem It Solves
OpenClaw runs locally. That is one of its defining design principles — your data stays on your machine, your agent operates in your workspace. The trade-off is that “locally” means you need to be at your computer to interact with it. For workflows that involve checking on a long-running task, asking a quick question, or receiving a notification when something finishes, the requirement to be at your desk is a real limitation.
The Telegram skill removes that constraint. Once installed and configured with a Telegram bot token, your agent maintains a persistent connection to a bot that you control on your phone. You send a message from Telegram, the agent receives it, processes it with full access to your local OpenClaw workspace and any other installed skills, and sends the response back to Telegram. The round trip typically takes two to five seconds. From a user perspective, it feels like messaging a very capable assistant.
Setup
Installation is the standard ClawHub one-liner:
npx clawhub@latest install telegramConfiguration requires a Telegram bot token, which you create through Telegram’s BotFather in about two minutes — it is one of the more painless API credential flows in existence. You message BotFather, run /newbot, pick a name, and receive a token. That token goes into the skill’s configuration, and from that point your agent is reachable on Telegram.
The skill also handles the reverse direction. Any OpenClaw skill or workflow that produces output can push notifications to your Telegram chat. Completed tasks, errors, summaries, alerts — anything you configure the agent to surface. This turns your phone into a passive monitoring interface for whatever is running on your machine.
What Changes When You Have It
The most immediate change is how you interact with your agent during the day. Instead of opening a terminal or switching to a desktop window, you message it the same way you message anyone else. Questions that are quick to ask but would otherwise require context-switching — “what did I have scheduled this afternoon?”, “did that script finish?”, “summarize the last email from this client” — become frictionless because the medium is already in your hand.
The second change is background task monitoring. OpenClaw agents can run multi-step processes that take minutes or longer — crawls, data processing, research workflows, builds. Without a notification mechanism, completing one of these requires either waiting at the terminal or periodically checking back. With the Telegram skill, the agent sends you a message when it finishes. You go do something else, your phone buzzes, and you have results.
The third change is mobile-initiated workflows. Because Telegram is a full input channel, not just an output one, you can start things from your phone that execute on your machine. Ask the agent to pull together a briefing before you get to the office. Kick off a research task while you are in transit. Request a file summary before a meeting. The agent runs on your hardware with access to your files; Telegram is just the interface through which you reach it.
Combination Installs
The Telegram skill appears frequently alongside skills that benefit from push notifications. The Mission Control skill — which generates morning briefings by aggregating mail, calendar, and task data — becomes more useful when it can push that briefing to your phone rather than waiting for you to ask. The Capability Evolver skill, which runs background self-improvement cycles, can notify you when it completes an improvement or encounters a constraint. Any long-running skill is a candidate for pairing with Telegram notifications.
There is also an emerging pattern of using Telegram as a lightweight approval interface. Workflows that require a human decision before proceeding — “I found 14 emails that look like they need responses, should I draft replies?” — can surface the decision to Telegram and wait for a reply before continuing. This fits naturally into agentic workflows that are mostly automated but have deliberate checkpoints.
Why Telegram Specifically
Telegram’s bot API is well-designed, free to use, and has no rate limiting that matters for personal agent workflows. The BotFather setup takes two minutes rather than the half-day required to navigate some other messaging platform APIs. Telegram also has no practical restriction on message length, supports code-formatted responses natively, and works reliably across mobile and desktop. For the specific use case of a personal AI agent notification and input channel, it is the most practical option currently available — which is probably why the skill has the download numbers it does.
The WhatsApp skill (wacli) covers similar ground for users already embedded in WhatsApp, and a Slack skill exists for team notification workflows. But Telegram remains the default recommendation in the OpenClaw community for personal agent access, and the install numbers reflect that consensus.
The Takeaway
145,000 installs is not an accident. The Telegram skill solves a concrete problem — local-first agents are powerful but desktop-bound — in the simplest possible way. If you run OpenClaw and you are not already using it, the setup takes about five minutes and the change to how you interact with your agent is immediate.
Leave a Reply