B Buffy Agent
Buffy Agent Blog · Interfaces

OpenClaw → Telegram Habit Agent: Plan in ChatGPT, Execute on Mobile

How to use OpenClaw with Buffy so your habit agent is designed for execution in Telegram—fast nudges, quick completions, and one shared behavior core.

Most people design habits where they think (often in ChatGPT), but they execute habits where they are reachable—usually on mobile.

That’s why the OpenClaw → Telegram workflow works: plan in ChatGPT, execute in Telegram, with one behavior core (Buffy) in the middle. This post gives you a sample reminder thread, a setup checklist, and how to avoid the trap of a siloed “Telegram habit bot.”

What you’ll learn

  • Why Telegram should stay a thin execution surface.
  • A copy-paste-style DM thread (including a skip path).
  • A 5-step checklist plus a quick verification you can run today.

The pattern: define once, execute everywhere

The reliable flow looks like:

  1. Define habits and routines in a planning surface (ChatGPT/OpenClaw).
  2. Execute with nudges in Telegram.
  3. Review with a briefing (instead of another dashboard).

If you want the OpenClaw entry point:

Why Telegram is the best execution channel

Telegram works because:

  • A “done” reply is fast.
  • Nudges are visible when you’re away from your laptop.
  • Habit follow-through often depends on mobile moments (walks, water, stretch).

The key requirement is that Telegram must not become a second source of truth.

Buffy keeps Telegram thin

Buffy treats Telegram as an interface adapter:

  • Telegram receives messages.
  • Buffy core stores activities and history.
  • Buffy’s reminder engine decides pacing and follow-ups.

That’s how you get a Telegram habit agent without duplicating logic across bots.

Related:

Sample reminder thread (what you see in Telegram)

A typical morning might look like this in your Telegram DM with Buffy:

Who Message
Buffy Morning startup window’s open (7:30–8:00). Water first—done now, or snooze 15m?
You done
Buffy Logged. Planning (10 min) and stretch still in the window. Do them now or snooze 20?
You snooze 20
Buffy I’ll nudge again in 20. Otherwise I’ll go quiet until your daily briefing.

If you skip the routine (honest data, not guilt):

Who Message
Buffy Last check in this window—morning startup. Do the remaining steps, or skip today?
You skip today
Buffy Logged. Want the same window tomorrow, or shift to 7:45–8:15 for this week?

All of that is driven by the same behavior core; Telegram is just the surface. No separate “Telegram habit bot” logic.

Setup checklist (OpenClaw → Telegram in 5 steps)

  1. Connect Buffy to your OpenClaw/ChatGPT workflow (one-time).
  2. Add Telegram as your execution channel in Buffy (link your Telegram account).
  3. Define one routine in ChatGPT (e.g. “Morning startup: water, plan, stretch” with a time window).
  4. Confirm Buffy will send reminders to Telegram (not only in-chat).
  5. Test with one “done” / “snooze” reply so you see the thread in Telegram.

Verify (30 seconds): In ChatGPT (or your planning surface), ask whether today’s routine shows the same state you just logged in Telegram. If the answer disagrees, fix adapter or account linking before adding more habits.

Boundary rules (adapter notes)

  • Telegram is a thin execution surface: send “done / snooze / skip” and receive reminders. The activity definitions belong in your behavior core.
  • Avoid creating a separate “Telegram checklist” concept. If the routine meaning changes, update the routine in ChatGPT/OpenClaw and let Buffy reflect it everywhere.
  • The fastest way to confirm you’re not siloing history is the verify loop: log once in Telegram, then ask for today’s status in ChatGPT.

After that, you can add more habits or routines; they’ll use the same pipeline.

Full Telegram connect steps (if you need screen-by-screen detail): How to Set Up Buffy in Telegram.

A day-in-life example

  • Morning: you define “Morning startup routine” in ChatGPT.
  • During commute: Telegram nudge appears (“window open—water now or snooze?”).
  • Midday: Telegram follow-up checks if you want to shift the stretch to later.

Because the behavior engine is shared, you can:

  • edit the routine in ChatGPT
  • complete steps in Telegram
  • coordinate team routines in Slack

FAQ

Do I need a separate Telegram habit tracker bot?

Not if your behavior core is shared. If you spin up a separate bot, you usually end up with fragmented history and inconsistent reminder behavior.

How do I stop reminders from becoming spam?

Use a reminder UX that offers exits (done/snooze/skip) and adapts with history:

Where to go next

Further reading