Day 6: OpenClaw Heartbeat, Cron Jobs & Memory — Proactive AI Assistant (2026)
OpenClaw Day 6: configure heartbeat checks, cron scheduled tasks, and three-layer memory (daily logs, MEMORY.md, SOUL.md) so your private AI assistant works while you sleep.
Part of the 7-day OpenClaw learning path on One Claw — SEO-friendly tutorials for hosted OpenClaw, private AI assistant, and OpenClaw skills.
Day 6 of the 7-day OpenClaw course upgrades your agent from on-demand tool to proactive co-pilot: heartbeat, cron, and the three-layer memory system.
Related: AI agent scheduled tasks on One Claw.
Heartbeat — the agent wakes up without you
Heartbeat is periodic polling: the agent checks inbox, calendar, monitors, or queues, then messages you only when needed.
Typical config areas (see upstream OpenClaw docs):
- Interval (e.g. every 30–60 minutes during work hours)
- Quiet hours (respect
USER.mdtimezone) - Channels to notify (Telegram vs workspace only)
Good heartbeat jobs:
- Important email since last check
- Calendar event in next 2 hours
- Error spike on a monitored URL
Bad heartbeat jobs:
- Re-sending full inbox dumps
- Nagging every 5 minutes
Cron — scheduled tasks with intent
Use cron (or One Claw Tasks) for time-based work:
| Schedule | Example |
|---|---|
| Daily 8:00 | Morning brief (Gmail + calendar + metrics) |
| Weekly Mon 9:00 | SEO / growth report |
| Hourly | Competitor pricing page diff |
Pair cron with skills from Day 4–5. Store outputs in memory files for continuity.
On One Claw, use the Tasks UI for team-visible schedules; gateway cron still runs the OpenClaw agent underneath on managed instances.
Three-layer memory
| Layer | File / mechanism | Purpose |
|---|---|---|
| Short-term | Today's chat + memory/YYYY-MM-DD.md | Running context |
| Long-term | MEMORY.md | Stable preferences, projects, decisions |
| Identity | SOUL.md (+ USER.md, AGENTS.md) | Who agent & user are |
Hygiene rules:
- Append — don't wipe
MEMORY.mdcasually - Promote only facts you want remembered months later
- Review weekly for stale entries
From passive to proactive — example day
- 08:00 — Cron morning brief to Telegram
- 10:30 — You ask ad-hoc question; agent uses today's memory file
- 14:00 — Heartbeat warns about calendar conflict
- 21:00 — Agent writes daily log + updates
MEMORY.md
This is the loop described in Day 1: Meet OpenClaw.
Safety for proactive agents
- Confirm before send email, post message, or spend money
- Rate-limit external API calls
- Log heartbeat actions for audit
Day 6 checklist
- At least one heartbeat or cron job enabled
- Morning brief or equivalent delivered once
-
memory/YYYY-MM-DD.mdcreated automatically or manually -
MEMORY.mdupdated with one long-term fact
Key takeaways
- Heartbeat + cron = assistant works while you sleep
- Memory layers prevent amnesia between sessions
- Day 7: advanced builds, security, and what comes next
Previous & next
One Claw resources
More Posts
Prompt words are copied and pasted everywhere, why teams need an AI skill library more than more dialogue windows
Introducing the value of One Claw skills management to help teams turn high-frequency prompt words, standard processes and best practices into reusable assets.
Day 1: Meet OpenClaw — AI Assistant vs Chatbot (2026 Guide)
What is OpenClaw? Learn how a private AI assistant differs from ChatGPT-style chatbots — proactive heartbeat, skills, memory, Telegram/Discord channels, and how to start on One Claw in minutes.
What Are AI Scheduled Tasks Good For? Turn Repetitive Work Into Reliable Routines
Use One Claw tasks and scheduling for daily reports, follow-ups, content drafts, and monitoring alerts—not just a demo feature.
Newsletter
Waitlist
Subscribe to our newsletter for the latest news and updates