Playbooks & routines
Run pre-built, tier-gated workflows from the Routines tab — the current superset of Playbooks.
AI
Preview — early access.
Overview
The Routines tab is a catalog of multi-step workflows grouped by category — governance, fleet health, site ops, content, business, and personal. Built-in routines are always on; everything else has to be activated before it can run. Routines is the current, broader name for what used to be called Playbooks.
How it works
Each routine is a named sequence of steps, and each step carries a risk tier (1–3). Tier-2-and-up steps ask you to confirm before they run. Steps run in order — a step stays locked until the one before it succeeds. Every run streams into the workspace's event log as ide.workflow.*, and a routine with a cron trigger shows a copyable /schedule line so you can hand it to the scheduler.
Steps
- •
Open the Routines tab
It sits alongside Browser, Secrets, Models, Git, Chats, and Terminal in the IDE's tab strip.
- •
Check what's already active
Built-in routines are always on and listed first; anything else needs activating.
- •
Activate a routine
Toggle Activate on a marketplace routine you want available.
- •
Run a step
Click "Run now" (or "Run next" once you're past the first step). Tier-2+ steps ask you to confirm before they run.
- •
Schedule it, if it has a cron trigger
Copy the
/scheduleline shown on the routine and hand it to the scheduler for a recurring run.
Capabilities
- Category-grouped catalog — governance, fleet health, site ops, content, business, personal.
- Built-in vs. marketplace — always-on core routines plus an activatable marketplace of others.
- Risk-tier gating — tier-2+ steps require an explicit confirmation before they run.
- Inline run output — each step's pass/fail result and output render right there in the card.
- One-click scheduling — a copyable
/schedulecommand for any routine with a cron trigger.
Limits & honest scope
- Some steps are marked "Coming soon" — the routine is defined, but the script behind that specific step isn't wired yet.
- Steps are strictly sequential; you can't skip ahead to a later step before an earlier one succeeds.
- Activation state is per workspace — activating a routine doesn't turn it on everywhere you work.