Skip to main content
Routines and goals turn your AI team into a proactive operation. A routine is a recurring automation your agents run on a schedule. A goal is a workspace-level objective the agents keep in mind and work toward over time. Together they let the team act without waiting for a new request each time.

Routines

Recurring automations, the team runs the same task on a schedule and reports back.

Goals

Standing objectives, the team weighs its work against what you’re trying to achieve.

​
What a routine is

A routine is a saved instruction the agents execute automatically at a set interval, daily, weekly, or on a custom schedule. Instead of asking “review the open issues” every morning, you define it once and the team runs it for you. Common routines:
  • A daily code-health sweep that flags risky changes and stale branches.
  • A weekly dependency check that opens issues for outdated packages.
  • A recurring review pass over recently merged work.
  • A scheduled knowledge-base refresh after a busy week.
Routines run locally like any other request. Premium models plan the work; the bundled local Forge executes code changes at $0 API cost.

​
Set up a routine

Open routines

Go to your workspace and open the Routines panel.

Describe the task

Write the instruction in plain language, just like a normal request, for example, “Check for failing tests and open an issue for each one.”

Pick a schedule

Choose how often it runs, daily, weekly, or a custom cadence, and the time it should start.

Save and let it run

Save the routine. The team runs it automatically on schedule and reports results back as trackable work.
Keep each routine focused on one outcome. A tight routine produces a clear report; a broad one produces noise. Use multiple routines instead of one catch-all.

​
What a workspace goal is

A workspace goal is a standing objective you set for the whole workspace, the bigger thing you’re trying to achieve. Goals don’t run on a schedule. Instead, the agents keep them in context and let them shape how they plan and prioritize work. Examples of goals:
  • “Ship the v2 dashboard by end of quarter.”
  • “Reduce the backlog of open bugs.”
  • “Improve test coverage across the API.”

​
How the agents act on routines and goals

The CEO orchestrator reads your repos, then plans and delegates to specialists, Tech Lead, Code Reviewer, Frontend, Backend, DevOps, QA, and Designer. Routines and goals feed directly into that planning.
When a routine is due, the orchestrator picks it up, plans the task, and delegates it to the right specialists, the same flow as a request you type yourself. Each run produces issues, code changes, reviews, and knowledge-base entries you can track.
Goals don’t fire on a clock. The team weighs new and recurring work against your active goals, so the changes it proposes move you toward the objective rather than drifting away from it.
Routine runs and goal-driven work land in ~/.orkestral as issues, reviews, and history, no server, no telemetry. You stay in control and can review every result.
Routines act on your codebase automatically. Review what each routine produces before relying on it, and keep schedules conservative until you trust the output.

Issues & work tracking

See how routine runs and goal-driven work become trackable issues.

Your AI team

Meet the specialists the orchestrator delegates to.