GearScheduled tasks

Run tasks on autopilot

Write the prompt once, pick a schedule, and an agent runs it — every night, every Monday, or on your own cron. Each run ends as a diff you can review and ship.

Construction capy

What to automate

The four schedules teams set up first — maintenance that used to eat sprint time, now running unattended.

Nightly test suite

The full suite runs overnight. Failures get investigated and fixed by an agent, so mornings start green instead of with a broken main.

Capy under a night sky

Weekly dependency updates

Patch and minor versions get bumped, tests run against the result, and a summary of what moved lands with the diff. Staying current stops being a chore ticket.

A stacked database

Security reviews

Auth flows, API endpoints, and database queries get scanned on a recurring schedule. Findings arrive as GitHub issues, ready to triage.

A computer with a shield on screen

Daily cleanup sweeps

Lint runs with auto-fix before your team logs on, and dead code, unused imports, and redundant logic get swept out on schedule. Style nits and cruft stop reaching review.

A clean brush stroke
Scientist capy

How it works

Three steps. No infrastructure. No cron servers to manage.

1

Create

A name and a prompt, written the way you'd brief any task. The prompt is the spec — the agent follows it on every run.

2

Schedule

Pick a preset — daily, weekdays, weekly — or write your own cron expression. Set the timezone and choose which model runs it.

3

Review

Each run lands as a normal task. Inspect the diff, open a PR, or send feedback — the output is never a black box.

Cron-deep control

Presets cover the common cases; raw cron expressions, per-automation branches, and model choice cover everything else.

  • Presets or raw cron expressions
  • Build for coding, Captain for planning
  • Any model available on Capy
  • Timezone-aware schedules
  • Pause without deleting
  • Run history with per-run status
  • Every run is a reviewable task
  • Base branch set per automation
Capy holding the key

FAQ

Schedules, models, failed runs, and pausing — answered.

  • How is an automation different from a regular task?

    An automation is a task that runs on a schedule. Each time it triggers, Capy creates a new task with your prompt, on the branch you configured. You review the output exactly like any other task: check the diff, create a PR, send feedback.

  • What schedule options are available?

    You can choose from presets (every day, every weekday, every week on a specific day, every N hours, every N minutes) or write a custom cron expression for full control. All schedules respect the timezone you set.

  • Can I use Captain for automations?

    Yes. You can choose between Build (for direct coding tasks) and Captain (for planning and delegation). Captain is great for recurring analysis, architecture reviews, or multi-step workflows.

  • What happens if an automation run fails?

    Failed runs appear in the run history with their status. You can review what went wrong, adjust the prompt, and the next scheduled run will try again with your updated instructions.

  • Can I pause an automation temporarily?

    Yes, just toggle the automation off without deleting it. All configuration is preserved. Toggle it back on whenever you're ready.

Set it and ship it

A prompt and a schedule — your first automation takes under a minute to set up.

Capys jumping through a portal