Loading...
Loading...
Run tasks on autopilot. Morning briefings, health checks, data syncs — set them once and your agent handles them without manual intervention.
Scheduled commands are agent tasks that run automatically on a time interval — like cron jobs, but instead of shell scripts, they trigger full Claude Code sessions. Your agent reads the command definition, executes the work, and logs the result.
They work best for recurring workflows with predictable inputs: pulling data, generating reports, running checks, archiving logs, or sending briefings. Any task you find yourself triggering manually every day is a candidate for scheduling.
Scheduled commands only work when your agent runtime is always on. A laptop that sleeps won't fire 3 AM tasks. This is one of the primary reasons to run agents on a VPS or use a hosted agent service.
These are the most frequently configured schedules across production agent setups.
0 7 * * *Summarize yesterday, surface priorities, check pending items
0 21 * * *Archive session notes, update memory, plan next day
0 */6 * * *Verify agent responsiveness and system integrity
30 8 * * *Pull external data sources, refresh dashboards
0 9 * * 1Retrospective, OKR check-in, priority reset
0 0 * * 0Archive old session logs, compress audit trails
Create a .md file in .claude/commands/ describing what the agent should do when the schedule fires.
Register the command in .claude/settings.json under the "schedules" key with a cron expression.
Point the schedule at your agent runtime (claude-code --dangerously-skip-permissions in headless mode, or a hosted agent).
Run the command once by hand to confirm output before enabling the schedule.
Scheduled runs append to .claude/logs/scheduled.log — tail it to verify the schedule fires correctly.
Add schedules to your .claude/settings.json:
{
"schedules": [
{
"name": "morning-briefing",
"command": "/morning",
"cron": "0 7 * * *",
"enabled": true
},
{
"name": "health-check",
"command": "/health",
"cron": "0 */6 * * *",
"enabled": true
}
]
}The command field maps to a file in .claude/commands/. The agent reads the file and executes the instructions as a full session.
Every scheduled run is logged. Check execution history with:
# Tail the scheduled task log tail -f .claude/logs/scheduled.log # View last 20 runs tail -n 40 .claude/logs/scheduled.log # Check for failures only grep "FAILED" .claude/logs/scheduled.log
Every managed agent plan comes with morning briefings, health checks, and daily syncs pre-configured and running on day one.
View Pricing