Developer
Start an agent when your system says so.
Send an event and an agent runs. Subscribe to platform events and your systems learn what agents did — runs started and finished, approvals requested and decided, budgets reached, agents paused.
What connects
Signed HTTP webhooks, in both directions.
What crosses the boundary
- Into the platform: your event payload, treated as data and never as instruction.
- Out of the platform: run lifecycle, approval and budget events, signed.
- Never: secrets, credentials or another organisation's events.
The capabilities this exposes
- Trigger an agent run
- Subscribe to platform events
Setting it up
- 1Register an endpoint and store its signing secret.
- 2Choose which events you want.
- 3Point an agent's trigger at your event type.
- 4Verify signatures on every delivery.
Permissions
Inbound events are authenticated and scoped to your organisation. Outbound deliveries are signed so you can verify they came from us. An inbound event can start an agent and cannot change what that agent is permitted to do.
Jobs an agent can own once this is connected
- Run an agent when an order is flagged in your system
- Notify your systems when an approval is granted
- Alert your on-call when an agent pauses on a budget
Where it stops
An event triggers an agent; it does not grant one anything. A payload cannot widen permissions, change a boundary or alter an approval rule — it is data the agent reads. Delivery is at-least-once, so your handlers need to tolerate a duplicate. Events are also not instant: this platform is built for work measured in minutes, and an agent is the wrong tool for something needing sub-second response.
Developer
Others in this category.
Give AI a job.
Build an agent, choose what it can access, define what it may do, and deploy it to work safely across your business.
No code required · Full SDKs and API · Stop any agent instantly · [email protected]