Launch your first agent in about five minutes.

Create a workspace

Sign up with your work email. Workspaces separate projects and control who can see what.

If you're on the free tier you can skip payment and start immediately.

Pick a template

Open the templates browser and find one that matches what you want done. Most first-time users start with "inbound lead triage," "meeting prep brief," "inbox zero assistant," or "weekly exec report."

Click the template name to see its skill composition, trigger shape, and approval defaults.

Customize what's specific to you

The template asks for the few details only you can provide. A Slack channel. A contact list. An approval email. A budget ceiling.

Skills that need credentials show an "authorize" button. Click it, the workspace walks you through OAuth or API-key entry, and the credential lands in the vault.

Set the cadence

Tell the agent when to act: on a schedule, on a webhook, on a human trigger. The builder shows the next three scheduled fires so you can sanity-check.

Approve a dry-run

The agent does a full dry-run without touching the outside world. You see exactly what it would have sent, written, or changed.

Approve if it looks right. Edit and re-run if it doesn't. Nothing is live yet.

Go live at L1

The agent is now live at trust tier L1: it queues every outbound action for your approval. You tap approve per action. Most teams run this way for a week before promoting.

Promote when you're ready

Once you've seen the agent work, promote it one tier (to L2 or L3) so it can act without per-action clicks. You can always demote again. Read the trust model for the full picture.