What lifecycle automations are
Lifecycle automation is the work that happens around an agent once it exists — the upkeep, recurring runs, and safety checks that keep it useful without someone starting a chat every time. Where the Agent Builder defines what an agent is, lifecycle automation defines when and how it runs on its own.
The three building blocks
| Mechanism | What it does | Where it lives |
|---|---|---|
| Scheduled & recurring runs | Run an agent at a set time or on a repeating schedule. | /automation — managed as tasks. |
| Event triggers | Run an agent when an external event happens (a new email, ticket, record, etc.). | Triggers tab in the builder. |
| Guardrails (before/after checks) | Validate, redact, or block input before the model and output after it. | Capabilities → Guardrails. |
Together these let an agent operate as a background worker: it wakes on a schedule or an event, runs its instructions, and is wrapped by safety checks each time.
When to use it
Reach for lifecycle automation when an agent needs to:
- Produce something on a cadence — a daily digest, a weekly report, a monthly refresh.
- React to events instead of waiting for a person to ask.
- Re-check or refresh data so its answers stay current.
- Apply consistent safety or policy checks on every run.
If an agent only ever runs when a person chats with it, you don’t need lifecycle automation yet.
How the pages in this section fit
- Pre/post hooks — the before-and-after checks that wrap each run.
- Scheduled maintenance runs — recurring jobs that keep an agent and its data healthy.