Skip to main content
Automations are rules that spawn coding agents in response to events. An automation doesn’t do the work itself — it spins up a coding agent in an environment. Triggers decide when to run, instructions and permissions decide what the agent is allowed to do. See Agents. Think of it like this: a trigger is the doorbell, the automation is the rulebook, and the agent is the person who answers and handles the task.

​
Three parts

An automation has three parts:
  1. Trigger — What event starts it (Sentry alert, Linear ticket, GitHub issue)
  2. Instructions — What the agent should do
  3. Permissions — Which actions are allowed (comment, open PR, update ticket)

​
Example

Imagine a Sentry error happens in production. Your automation launches an agent that investigates the stack trace, identifies a likely cause, and then reports back with a fix proposal.

​
What happens

When a matching Sentry error comes in, Proliferate:
1

Environment spins up

The agent starts in a real environment with your repo and dependencies.
2

Context is gathered

The agent receives the right details from Sentry, Linear, GitHub, or other tools.
3

Work is done

The agent investigates, makes changes, and prepares results.
4

Results are delivered

Outcomes show up in the client where the request started.