Runs and approvals

A run is one execution of a workflow, started from one trigger. Each node it reaches is a step, with a status, an input and an output. Every run counts once against your plan’s monthly runs, draft runs included. See Allowances.

Find a run

• In the builder, the Runs tab lists the latest runs. Select one to open it on the canvas.
• On the run history page, from More then Full history in the builder: every run, newest first, with its status, when it started, its version or Draft, and who or what started it.
• On the Dashboard: recent runs across every project, approvals waiting, stopped triggers and steps that failed in the last 7 days.

A run’s address is the builder’s with ?run=<run id> added, so you can share a link to it.

Run statuses

Status	What it means
queued	About to start
running	Steps are running
waiting	Asleep, such as during a Delay
complete	Every step that could run has finished
errored	A step failed and the run stopped
terminated	The run was stopped before it finished
unknown	The run is older than 30 days; its steps are still shown

Read a run

With a run open, each node on the canvas shows its step’s status and duration, and the edges that fired are drawn in its color. The panel lists the steps in run order. Select a step, or its node, to see:

• Input: the text of the step that fired into it.
• Each output field, labelled nodes.<id>.<field> as a template reads it.
• Error, when the step failed.

Under the steps, Reply is the text of the step or steps the run ended on.

Step status	What it means
running	In progress
waiting	Waiting for a member to approve or deny an action
done	Finished
error	Failed. The run stops.
skipped	Another trigger started the run, or no edge into the node fired

The run page shows at most 8,000 characters of each value. Templates in later steps read the whole value.

How steps run

• A step starts once every step before it has finished or been skipped. Branches that don’t depend on each other run in parallel.
• Every step is checkpointed. If the platform restarts mid-run, the run carries on from its last finished step.
• A step that fails isn’t retried, and the run stops with an error. The one exception is an email Notification with Retry if sending fails on.
• Several edges may lead into one node, but only one may fire in a run. If two fire, that step fails with an error naming both steps.

When a run fails

1. Open the run and select the step marked error.
2. Read its Error and Input.
3. Fix the node in the draft, then test it with Run draft and the same payload.
4. Publish once it works.

The assistant beside the builder sees the failed step and its error, and can help you fix it.

Approvals

An agent can be granted actions that ask first, and destructive actions always do. When an Agent step’s agent calls one, the step waits:

1. Every owner and admin with a verified email is emailed, with a link to the run and one to the agent’s thread.
2. The run page shows the waiting call: the action and its input, with Approve, Deny and an optional Reason. The Dashboard lists it under Waiting for approval.
3. Any member of the organization decides, on the run page or in the agent’s thread.
4. Approving runs the action. Denying tells the agent the reason. Either way the agent carries on, and the run continues.

A call left for 72 hours is denied with the reason “expired without a decision”, and the step fails. See Grants and approvals and Agents in workflows.

At your monthly run limit

When the month’s runs are used up:

• Run draft and API runs are refused with a message that the limit is reached.
• A scheduled time is skipped. The schedule continues next time.
• An app event is recorded as limit reached and starts no run.

Runs count again from the first day of the next month, in UTC. See Allowances.