← All topics

Automations

Run an agent on a schedule or when something happens — one screen for every trigger.

Why it works this way

Recurring and event-driven work should not depend on you remembering to open the thread. An automation is one named thing: when it fires, Nested starts a task with the name you gave it, and every run lands in its history.

Before you start

Steps

  1. Create an automation: pick the agent, give it a name (each run opens a task with this name), and build the When.
  2. Pick a trigger. Schedule (weekdays or interval), Webhook (HTTP method, path slugs, one-time HMAC secret), GitHub (events plus optional @mention filters), Slack (mentions, channel messages, emoji reactions), or a schema provider such as Sentry.
  3. Choose what the agent does: Instructions (markdown, with the event payload appended) or a Formula from the agent’s pack. The formula form is that formula’s input. A value you leave empty can be filled from the trigger event when the input declares that property.
  4. Optionally mark the task complete when the agent finishes, let the agent read a Slack channel while it runs, or post the final message back to Slack or the pull request.
  5. Save and turn on — or Save as paused and switch it on later from the detail page.
  6. On the detail page: pause, Run now, edit, delete, and review History — every run shows its trigger, status, and payload JSON, with Open task and Repeat.
  7. Stuck? Describe the automation in plain words and Fill the form, or Start from a template.

Example

When a pull request opens in acme/app → run the code-review agent with instructions “Review for bugs and comment.” Nested starts a task, the agent comments on the PR, and the run appears in History with its payload.

Open in Nested