What an agent is
A brief + model + tools, versioned, that anyone in the workspace can run.
A saved brief, model and set of knowledge bases that anyone in the workspace can run. Each edit publishes a version, and every run records the one it ran.
Chat answers one question. An agent does a job: searches several times, reads what it finds, calls other systems, writes files, and can run with nobody typing — on a schedule, from a webhook, through the API or from an embedded chat on your website.
Lifecycle
| Status | Meaning |
|---|---|
| draft | Just created or edited; nothing outside can run it yet. You can still test it in the Run tab. |
| active | After Activate. Triggers, the API and widgets may call it. |
| archived | Hidden from the list, does not run. |
Every Publish new version in Configuration or Flow bumps the version (v1, v2, …). Every
run records the version it used, so you always know which configuration produced a result — and
can go back to an old version at any time.
Plan limits
Free 2 agents, Starter 5, Pro 20, Team and up unlimited. Triggers and webhooks need a plan with Webhooks and triggers (Starter and up); API keys need Pro and up. See Plans and features.
In this section
Create an agent
From a template or from nothing.
Configure
Instructions, model, tools, approvals, memory.
Run
The Run tab, the timeline, mid-run approvals.
Flow
Draw a multi-step process on the canvas.
Triggers
Run on a schedule or when another system calls.
Versions
Compare, diff, go back.
History
Past runs and per-step detail.