Appearance
Scheduled Messages & Automation
Antigravity can run agents on a schedule. A morning PR review, an overnight dependency audit, a weekly changelog — anything you would otherwise run manually can be automated with scheduled messages. These run on Antigravity's infrastructure, so they keep running even when your machine is off.

What you'll learn
- Scheduled messages run agents on a recurring schedule — morning, daily, weekly, or custom cron
- They run on Antigravity's infrastructure, not your local machine
- Common automations: PR reviews, dependency audits, changelog generation, documentation sync
Build it
Step 1: Create a scheduled message
In the Antigravity IDE or CLI:
/schedule "Every weekday at 9 AM: Review all open PRs in the myorg/myproject repo. Summarize new PRs, flag any that need urgent attention, and post the summary to the #engineering channel."Step 2: Manage schedules
Control the lifecycle of your recurring background jobs using dedicated scheduling commands. You can list, pause, resume, or entirely remove active automations as project needs evolve.
bash
# List all scheduled messages
/schedules
# Pause a schedule
/schedule pause "morning-pr-review"
# Resume
/schedule resume "morning-pr-review"
# Remove
/schedule remove "morning-pr-review"Step 3: Common automation patterns
Use these templates to set up regular reviews of continuous integration pipelines, dependency management, and documentation synchronization. Adjust the schedule frequency and prompts to suit your specific team requirements.
bash
# Overnight CI failure analysis
"Every morning at 7 AM: Check the CI pipeline for failures in the last 24 hours. For each failure, analyze the logs, determine if it's a flake or a real issue, and post findings."
# Weekly dependency audit
"Every Monday at 10 AM: Check for outdated dependencies. For each major version bump, read the changelog and assess breaking changes. Create a PR for safe updates."
# Documentation sync
"After every merge to main: Check if any API endpoints changed. If they did, update the API documentation and open a PR."What goes wrong
| Mistake | How you notice it | The fix |
|---|---|---|
| Schedule fires but agent produces wrong output | PR review is generic or misses issues | Review the prompt. Add specific instructions and output format requirements |
| Schedule doesn't fire | No output at expected time | Check /schedules for status. Verify the schedule syntax. Check if the agent has necessary API access |
| Too many scheduled tasks overwhelm context | Agent output quality degrades over time | Limit to 3-5 scheduled tasks. Combine related tasks into one schedule |
Next: Multi-Workspace Projects