Module 6: Autonomous Tasks

Why this matters

So far, your agent only does things when you message it. In this module, you make it proactive — running tasks on a schedule and staying online around the clock.

Scheduling tasks with cron

OpenClaw has built-in cron support. You can schedule any prompt to run automatically.

Add a cron job

In your Terminal, schedule a task like this:

openclaw cron add "0 7 * * *" --message "Generate my daily briefing and send it to me"

This runs every day at 7:00 AM. You don't need to understand the scheduling syntax — just copy from the table below and change the time.

Common schedules

Manage cron jobs

# List all scheduled jobs
openclaw cron list

# Remove a job
openclaw cron remove [job-id]

See the CLI Cheatsheet for the full list of openclaw cron flags and options.

Agent Execution & Progress Tracking

When your agent runs automated tasks, it can report its progress with the /exec command.

Understanding execution reporting

As of v2026.4.8, /exec reports now accurately reflect your actual runtime environment. The fallback policy shown depends on where OpenClaw is running:

If you set host=auto, the policy automatically adapts to your environment. Previously, stricter defaults were displayed even when the actual behavior was more permissive. Now the report is honest about what your agent can do.

Progress tool: keeping update_plan available

For OpenAI-family runs, the agent keeps update_plan available for progress tracking during autonomous tasks. You can now:

• Set tools.experimental.planTool=true to enable progress reporting (default)
• Set tools.experimental.planTool=false to opt out and use simpler compact payloads

This gives you fine-grained control over whether the agent tracks detailed progress or sends minimal status updates.

Example: disable detailed planning for cost-sensitive automations

{
  "tools": {
    "experimental": {
      "planTool": false
    }
  }
}

Deployment: Keeping your agent online

Your agent currently runs on your computer. If your laptop sleeps or shuts down, the agent stops.

Choosing your deployment option

Before setting up 24/7 operation, pick the approach that fits your situation:

If you just want to test automations, they already work — your machine is the server. Only set up a VPS if you need guaranteed uptime.

Option 1 — Keep on your machine (works now)

Your cron jobs already run whenever the Gateway is running. If you're happy with automations firing only while your computer is on and awake, you're done with this section.

Option 2 — Prevent your machine from sleeping

On macOS, prevent sleep while OpenClaw is running:

caffeinate -d &

On Windows: Settings → System → Power → Screen and sleep → set "When plugged in, put my device to sleep after" to Never.

This gives you near-24/7 operation without any server setup, as long as your machine stays powered on.

Option 3 — Deploy to a VPS (most reliable)

A small cloud server costs $5-6/month and runs 24/7.

ssh root@YOUR_SERVER_IP

curl -fsSL https://deb.nodesource.com/setup_24.x | bash -

apt-get install -y nodejs

npm install -g openclaw@latest

openclaw onboard --install-daemon

scp ~/.openclaw/SOUL.md root@YOUR_SERVER_IP:~/.openclaw/
scp ~/.openclaw/USER.md root@YOUR_SERVER_IP:~/.openclaw/

openclaw gateway status

Home server (no monthly cost)

If you have a Raspberry Pi, old laptop, or NAS, you can run OpenClaw on it. Same setup as VPS, but the machine stays in your home. Your data never leaves your network.

Syncing configuration

When you update your SOUL.md or config locally and want to push changes to your server, run this from your local Terminal:

rsync -avz ~/.openclaw/ root@YOUR_SERVER_IP:~/.openclaw/ --exclude workspace/

rsync copies only the changed files — it is much faster than copying everything. The --exclude workspace/ skips your local files so they don't overwrite the server's memory.

Then restart the gateway on the server:

ssh root@YOUR_SERVER_IP "openclaw gateway restart"

Monitoring

Check status

openclaw gateway status
openclaw cron list

View logs

openclaw logs --tail 50

Automated health check

Add a self-check cron job that pings your agent every 30 minutes and only alerts you if something is wrong. In your Terminal:

openclaw cron add "*/30 * * * *" --message "Self-check: confirm you are running and responding. Reply with 'ok' only." --suppress-if-ok

--suppress-if-ok means: only send the result to your channel if the agent fails to respond. If everything is fine, you hear nothing. If it breaks, you get a message.

This checks every 30 minutes. Since it's a short prompt with a short reply, the API cost is negligible.

Health endpoint

The Gateway exposes a health endpoint you can monitor externally:

curl http://localhost:18789/health

Security for deployed agents

When your agent runs on a server:

• The Gateway binds to 127.0.0.1:18789 by default (not publicly accessible)
• Never expose port 18789 directly to the internet
• Use Tailscale for remote access (see Module 8)
• Keep your config file permissions restricted: chmod 600 ~/.openclaw/openclaw.json