Skip to main content
Once an agent is enabled, it can run without anyone pressing a button. People reach it through chat, channels and the Agent API; automations run it on a schedule or when another system calls a webhook.

Enable or disable an agent

An agent runs only while it is Enabled. The status pill at the top of the agent page shows Enabled or Disabled; click it to switch.
  • Publishing from the Agent Builder enables the agent.
  • Disabling stops it everywhere: chat, API calls, public links, email, chat apps, rooms and automations. Building and Test run in the builder keep working. Turn it back on to resume.
Only the Admin and Builder roles can enable or disable an agent.

Automations

An automation is a saved invocation of the agent: a prompt, the triggers that fire it, and where its results go. Every firing is a real run of the agent, and each run starts a new conversation.

Create an automation

1

Open the Automations tab

On the agent page, open Automations and click New Automation.
2

Name it and write the prompt

Give it a Name (for example “Morning digest”) and describe the task under What should it do?, for example: “Summarize last week’s new leads and flag anything unusual.”
3

Add triggers

Under Triggers, click Add trigger and pick Scheduled or Webhook. Add as many schedules as you need, plus one webhook. With no triggers, the automation runs only when someone presses Run now.
4

Choose where results go

Open Advanced settings to pick Destinations, and optionally the Model, Run as identity and Max concurrent runs. The editor summarizes the choice as Results go to: ….
5

Save

Click Create. A webhook URL appears once the automation is saved.

Run now

Every automation row on the Automations tab has a Run now button that fires it once immediately. Run now is available to the Operator, Builder and Admin roles.

Schedules

A Scheduled trigger runs the automation at set times, in the Time zone you choose. Custom cron examples: A paused schedule shows This schedule is paused. with a Resume button.

Webhooks

A Webhook trigger gives the automation its own URL. Any system that can send an HTTP POST can run the agent: n8n, Make, Zapier, GitHub, meeting recorders, your own apps. The URL has this shape, and the editor shows it with a Copy URL button under How to call this Automation:

Protect the webhook

Under API key (Bearer token), click Generate to require callers to send Authorization: Bearer <key>. The key is shown once; DataGen stores only a hash, so save it immediately. Rotate issues a new key and stops the old one at once. Without a key, anyone with the URL can fire the webhook. Max calls per hour limits how often the webhook can fire.

Use the payload in the prompt

The JSON body you POST is available to the prompt. Write {{ payload.foo }} in What should it do? to insert the request’s foo field:

Call it and get the result

1

Trigger the run

The response is 202 Accepted with { executionId, statusUrl }. The run starts asynchronously.
2

Poll for the result

Poll every 1-2 seconds until status is final. The agent’s answer is in result.content:
Click Test fire in the editor to POST {} to the URL and confirm it is queued.
For a request-and-response conversation with an agent from your own code, use the Agent API instead of a webhook.

Some older automations use the one-shot repo run, which can commit and open pull requests (PR mode: Create PR, Auto Merge or Skip). The editor shows a notice on these with a Switch to conversation button. New automations always run as conversations.
  1. Check that the agent is Enabled.
  2. Check that Automation active is on and the schedule is not paused.
  3. Check the Time zone on the trigger.
  4. Look for the run in Agent Logs, filtered by the Scheduled (Cron) trigger.
  • 401: the webhook has an API key. Send Authorization: Bearer <key>, or Rotate to get a new key if you lost it.
  • 404: check the agent name and automation id in the URL, that Automation active is on, and that the automation has an active webhook trigger.
  • 429: the webhook hit Max calls per hour, or you have too many queued or running runs. Retry after the time in the response.

Next

Channels

Reach the agent from Slack, Teams, Discord, email, web chat and public links

Versions and Logs

Roll back a version and inspect every run