Skip to main content

Outgoing Webhook

Call your system when an event occurs during workflow execution.

note

You can add outgoing hook anywhere into your workflows if you want to track execution more precisely

Pre-requisite

A Workflow with an Outgoing Webhook function node properly configured.

Configuration

See platform reference: How to configure Webhook

Add your own system url to trigger the hook

Wiring the node

The node has two inputs and no output — it is the end of its branch.

Tracker uuidrequired

The identifier of the current execution, which the node forwards to your system. Wire it from the Monitoring / Tracker uuid output of your entry point node — without it the step fails.

Trigger

Optional. Wire it to decide when the call fires — the hook waits for this branch to reach it. Accepts a trigger, an integer or a plain text value.

Usage

Zparse sends a POST request to your url with the following JSON body:

Request body:
{
"hook_uuid": "2b747c9e-e411-4eff-a0ea-feba84d2f180",
"workflow_uuid": "6b81a3e0-c644-4a8a-8804-19edfe4a9433",
"tracker_uuid": "{TRACKER_UUID}"
}
  • hook_uuid identify which workflow node made the call (useful if you have multiple outgoing hook in the same workflow)
  • workflow_uuid identify which workflow the request originated from
  • tracker_uuid get information about current execution instance

The payload is deliberately minimal — it tells you that something happened and where, not what the data was. Call back the Monitoring API with the tracker_uuid to read the run's steps, and retrieve a response if you need the payload a responder produced.

Answer within one second

Zparse waits one second for your endpoint and then gives up. Acknowledge immediately — return 200 as soon as you have queued the notification, and do your own processing afterwards. A slow endpoint fails the step even though it received the call.

There is no retry: a failed notification is not sent again.

Delivery semantics

  • Any HTTP status counts as delivered. The step succeeds as long as your server answered at all — a 500 or a 404 is treated as success. Only a timeout, a DNS failure or a refused connection fails it.
  • A failed call fails the step, which stops that branch of the workflow. Put the hook on a side branch if the rest of the run should continue regardless.
  • No authentication is sent. Zparse adds no Authorization header and no signature, so protect the receiving url yourself — keep it unguessable, or embed a secret in the path or the query string and verify it on your side.
  • Local and private addresses are rejected when saving the node: a url containing 127.0.0.1 or 10.0.0. is refused, so your endpoint must be reachable from the internet.
  • An empty url is flagged as a configuration warning on the node in the editor.

Common patterns

  • Notify on completion — place the hook at the end of the workflow so your system knows the run finished and can pull the result.
  • Notify on failure — wire it on the failure branch of a step to get alerted only when something goes wrong.
  • Track progress on long runs — drop several hooks along the workflow; use hook_uuid on your side to tell which milestone was reached.