Skip to content

Let your agent work while you sleep

At the end your Running coach writes a morning briefing at 6:30 while your devices are off, and sends it to you on Telegram.

You need: Pro, a first agent, and a Telegram account.

Your agents keep working in the cloud while your devices are off (Pro, part of Cloud Sync). Inside Cloud Sync, a small private program, your agent home, holds your data. It sleeps for free and wakes at once, on a schedule or when you message it. See The agent home.

At 6:30 your cloud box, inside Cloud Sync, wakes and runs the agent. The agent sends a briefing to Telegram. Anything the agent sends or posts on its own waits for an Approve button in Telegram.6:30the scheduleyour devices are offCloud Sync (Pro)Cloud boxRunning coachreads your dataTelegramthe briefingApprovebefore it sends
Cloud Sync does the work. You only tap Approve.
  1. Link your Telegram to your account in the agent home settings. See Chat channels.

  2. Make an automation: a schedule, every day at 6:30. Add one step, run an agent: Running coach, with the task “Write my morning briefing: yesterday’s training and today’s plan. Send it to me on Telegram.” The agent sends the briefing itself, to your own linked chat. If the run reads a collection set to Ask me first, it waits for your OK, so keep those out of night-time jobs.

  3. Approve the automation once, in the app. It then runs each morning by itself.

  4. Go to sleep. At 6:30 the agent runs in the cloud and sends the briefing to Telegram.

  5. Reply in Telegram: “Move today’s run to tomorrow.” The agent changes the entry. It is in your history, and you can undo it.

  6. If the agent wants to send or post something on its own, Telegram shows an Approve button (by default). It works once, for that action only. Nothing goes out until you tap it. Buying something waits for you in the app.

Safe by design

  • What the agent reads (a web page, a message) is information, never an order. Reading never gives it new permissions.
  • The agent’s browser comes in two stages. Research (next) reads public pages only, with no logins and no forms. Acting (later) signs in and fills forms only on sites you approved.
  • Every change to your data is in the run history and can be undone. A message already sent cannot be unsent. A read-only mode and an off switch stop the agent at once.
  • What waits for your Approve, and the agent’s daily spending limit, are your AI safety settings. Standard is the default. Pick Strict if your agent reads many web pages at night.
  • If Cloud Sync was not running at 6:30, it catches up later. A missed run that would send a message asks you first.

Make it yours

  • When Research mode comes (next), let the agent find a race near you and add it to your Goals.
  • Add WhatsApp as a second channel when it arrives.
  • Give the agent a weekly job too: a Sunday review. See Make your first automation.