Automation

Make OpenClaw proactive with cron jobs, background tasks, Task Flow, hooks, standing orders, heartbeat behavior, and external webhooks.

Automation turns OpenClaw from a reply bot into an agent system that can wake up, schedule work, react to events, and run background tasks.

Automation Surfaces

Cron Jobs

Cron is for predictable schedules:

openclaw cron create "0 8 * * 1-5" \
  "Prepare my workday briefing: calendar, urgent email, weather, and open follow-ups." \
  --name "morning-briefing" \
  --session isolated

Manage jobs:

openclaw cron list
openclaw cron show <job-id>
openclaw cron remove <job-id>

Recent OpenClaw versions report cron state through the current storage layer rather than assuming a legacy jobs file. Prefer CLI commands over hand-editing runtime state.

Background Tasks

Background tasks track work that should continue beyond a single foreground turn. They are important for:

• isolated cron runs
• ACP harness sessions
• subagent work
• long CLI operations
• Task Flow steps

Use task status before assuming work failed:

openclaw tasks list
openclaw tasks show <taskId>

Task Flow

Task Flow is the orchestration layer above background tasks. Use it when a workflow has multiple steps, checkpoints, or recoverable progress.

Good Task Flow candidates:

• research then draft then review
• deploy then verify then report
• classify inbox then prepare replies then request approval
• run checks then summarize failures then open follow-ups

Hooks

Hooks respond to OpenClaw events. A hook can run on lifecycle events, commands, session changes, outbound delivery, or other configured triggers.

Use hooks for:

• formatting reply payloads
• usage footers
• Slack outbound notifications
• custom audit trails
• workflow-specific post-processing

Inspect hooks through:

openclaw hooks list
openclaw hooks info <name>

Webhooks

Webhooks let external systems wake OpenClaw:

• CI/CD notifications
• issue tracker events
• support queue events
• monitoring alerts
• Gmail Pub/Sub runners where configured

Keep webhook paths protected with tokens, bind locally when possible, and expose through a trusted reverse proxy or tunnel when remote access is required.

Standing Orders and Heartbeat

Standing orders define durable authority. Use them sparingly. A standing order should say what the agent may do, how often, with what data, and when it must ask.

Heartbeat is a periodic check-in. It is better for lightweight monitoring than exact schedules.

Example: Safe Email Triage

1. Install/authorize the mail capability.
2. Test read-only inbox search.
3. Create labels manually.
4. Schedule a cron job that classifies new messages.
5. Draft replies but require confirmation before sending.
6. Add a usage footer or summary hook if desired.

openclaw cron create "*/20 * * * *" \
  "Review new unread mail since the last run. Label only. Draft replies for action-required items, but do not send." \
  --name "email-triage" \
  --session isolated

Failure Modes

Watch for:

• cron jobs using stale credentials
• background tasks still running after the user assumes they stopped
• hooks that modify replies unexpectedly
• webhooks reachable without enough auth
• standing orders that are too broad
• automation using a powerful main session when an isolated session would be safer