Runs and replay: watch AI workflows run step by step

Press Run in the editor toolbar (⌘ ↵). The editor validates the draft first; if anything would stop it running, such as a missing credential or a template that points at a node that does not exist, the issues are listed above the canvas and nothing starts.

The run starts from the trigger with its test payload. While it runs, the same button cancels it. A cancelled run starts no new nodes, and anything unfinished is marked cancelled.

Each node shows its status as the run reaches it. The run panel lists the same thing as a table, and for an agent node, the text it is streaming back from Claude.

• Pending: waiting for its inputs.
• Running: executing now, retries included.
• Succeeded or Failed: finished, and which port fired.
• Skipped: every edge into it was on a branch that did not run, such as the false side of an If.
• Cancelled: the run was cancelled before this node finished.

Status arrives over a live event stream, so there is nothing to refresh. The engine design post explains how.

Click any node that has run to open the output inspector. It shows the node's Input, its Output, the Error if it failed, and for an agent the tokens it used. This is usually the fastest way to see why a branch went the way it did: the If node's input shows the exact value it compared.

Select a node and choose Test this node in its configuration panel. Worfilo runs only that node, against the upstream outputs from the last run, with the configuration on screen, saved or not. The result shows its status, how long it took, and the tokens it used.

Open the Runs tab in the editor to see every run of the workflow: its status, what started it (manual from the editor, schedule from a trigger, playground, or api from the workflow API), when it started, how long it took, and the tokens it used.

Open any run from the history to replay it on a read-only canvas. The canvas is drawn from the snapshot of the graph that actually ran, not the current draft, so an old run still makes sense after you rename nodes, change prompts or publish new versions.

Every node keeps its final status, and clicking one shows the input and output it had in that run. A failed run opens on the node that failed. Credentials and secrets are redacted before anything is stored, so a replay never shows a key. See Credentials and security.