Flows and runs

A flow names a source app and one HTTP destination. Each incoming event becomes a run with its own input, destination and delivery key. An owner reviews that snapshot before delivery.

Create, activate and pause a flow

In Flows (/app), fill in Create a flow and select Create draft flow. Owners, admins and agents can create drafts. Only an owner can select Activate flow. A flow must be active before it accepts a new event.

Select Pause flow to stop new events being admitted. A paused flow shows draft and can be activated again. Pausing does not cancel an approved request already on its way to the destination. Pending runs can still be approved or rejected deliberately by the owner.

For incoming events, use a Trigger credential scoped to that flow. To enter one by hand, expand Prepare a run on an active flow, supply an Event ID and an Event input (JSON object), then select Prepare for approval.

The event ID identifies an event within its flow. Submitting the same ID with the same input returns the existing run. Reusing it with different input is refused. An erased event cannot be admitted again under that flow and event ID.

Decide on a run

Open Runs (/app/runs), or select View runs on a flow. The list shows the newest 200 runs across the workspace; View runs filters that list. Select the event ID to open Run details.

Check the source, destination URL, flow revision, delivery key and Exact input sent on approval. For a pending run, an owner can select:

• Approve and deliver: record the decision and make the first HTTP delivery attempt.
• Reject: record the decision without sending anything.

The decision is recorded against the reviewed action. Approval history shows who decided and when. Delivery attempts shows each attempt and its outcome. Crank retains safe outcome details rather than arbitrary destination response content. A succeeded result confirms the destination's acknowledgement; it does not certify a later business action there.

Replay a failed delivery once

Crank does not retry a failed delivery on its own. Check the destination first: a timeout or interrupted request may have reached it even if Crank could not confirm receipt.

An owner can select Replay once after a failed first attempt. An interrupted attempt whose time limit has expired also becomes eligible. Replay sends the same approved body to the same destination with the same delivery key. It uses the run's only replay, even if that attempt also fails. Rejected and successful runs cannot be replayed.

Connect your own destination

Crank sends a JSON POST to a URL on the installation's single configured destination origin, CRANK_TARGET_ORIGIN. Outside local development the origin must use HTTPS. Destination URLs cannot contain credentials, query strings or fragments. Redirects are refused. Authentication for the destination, if needed, comes from the installation's CRANK_TARGET_TOKEN binding; it is not entered in a flow or imported from n8n.

The receiver must atomically store the Idempotency-Key header and body digest with its effect, refuse changed bytes under the same key, and acknowledge both first delivery and duplicates with JSON:

{"id":"destination-id","idempotencyKey":"the-run-id"}

The delivery key is the run ID shown in Run details. The receiver's handling of that key makes a replay safe; an arbitrary HTTP endpoint is not sufficient. See the README deployment instructions and HTTP API for configuration and request details.

Continue with Access, Move in or Backup, restore and erasure. For a local walkthrough, see Get started.