Skip to main content
The Trigger node decides when a refinery runs. Four kinds: Schedule, Watch, Webhook, API. Regardless of the trigger, the canvas [Run] button or the CLI (nora pipelines) can kick off a run anytime.
Schedule Trigger settings showing cadence, time, timezone, and a Seed input

​
Schedule

Runs at set times.
  • Preset: pick from daily / weekly / monthly / yearly, then time and interval.
  • Custom cron: write a 5-field cron. 0 9 * * 1 is every Monday at 09:00.
  • Timezone: defaults to UTC. Switch to Asia/Seoul for Korea time.
After you set it, the inspector shows the schedule in plain language (“Runs every day at 09:00 (UTC).”). A Seed input (optional) is passed as user_input on every run. Same field as in Workflow. Use it for per-run inputs like an RSS URL.

​
Watch

Watches a connected source and runs when something changes. Good for keeping knowledge in sync with a folder as new documents land.
  • Trigger on: New file, New or modified file, or Any change (including deletes).
  • Debounce: default 5 seconds. If several files arrive at once, they’re batched into a single run instead of one run per file.
Watch only supports file-shaped sources: File (Nora Library) · Cloud Source (Google Drive, etc.) · Local Filesystem. Database, Web, Object, and Stream sources cannot be watched. Attach a Schedule trigger for periodic polling instead.

​
Webhook

Turns the refinery into an HTTP endpoint. Any system that can POST JSON can trigger a run. Handy when your CMS, ticketing, or CI already produces webhooks.
  • Endpoint: the POST URL built from the webhook slug. Copy from the inspector.
  • Auth: header-token style. Default header is X-Webhook-Token; hit [Generate token] to have the server mint a random token. Send it on that header when calling.
  • Body mapping: after the first POST arrives, hit [Refresh] to see the body. Click fields to map them to input, image, audio, Thread ID, or User ID.

​
API

For calling from your own code. Where Webhook receives events from external systems, API is where your service directly invokes the refinery.
  • Endpoint: POST to …/api/v1/<workspace>/foundry/<slug>.
  • Body: input (object), variables (object), idempotency_key (string).
  • Auth: header token, default header Authorization. Token generation is the same as Webhook.

​
Trigger kinds and graph rules

Trigger nodes split into two families, and each family drives different graph requirements.
  • Payload triggers (Webhook · API): the caller hands the body in, so the trigger is the input. No Source node needed and no Output node needed either; the result is returned inline in the response body. Right for validation / notification / compute pipelines that just calculate a value and hand it back.
  • Signal triggers (Schedule · Watch): these only say when, so a Source node has to read the real data. Saving without a Source errors out, and because the run has to persist somewhere, at least one Output node is required too.
If a pipeline mixes both families, the Signal rules win. A Source and an Output are still required.

​
Common settings

The Advanced tab on any Trigger node exposes retry count, backoff, timeout, and the raw JSON. Locking the node blocks edits via MCP or CLI.