Claude Code /loop puts agents on schedule

The interesting part here is not “fully autonomous coding” but a much more practical shift: Claude Code is becoming a lightweight automation layer for recurring engineering work. That makes this more of a workflow tutorial than a flashy product launch.

• –`/loop` gives Claude Code cron-like recurring execution, including natural-language intervals for checks like deploy health, PR review, and error-log sweeps
• –The model is intentionally constrained: scheduled jobs are session-scoped, auto-expire after a few days, and are better for monitoring and repetitive prompts than long autonomous builds
• –Because scheduled prompts can invoke slash commands and skills, Claude Code starts to look like a developer-facing orchestration tool, not just a chatty CLI
• –The operational downside is cost and observability: aggressive loops can quietly burn usage unless paired with notifications, sane intervals, and log monitoring
• –For more durable unattended workflows, developers will still need external tooling like GitHub Actions or other schedulers instead of relying on terminal-bound sessions alone