Skip to main content

Running workflows

Press Run in the canvas top bar and Corte executes the saved graph in dependency order: inputs resolve first, then each generator fires as soon as everything it's wired to is ready. From the second run on, only what you changed is made again — everything else keeps its last result, free.

app.corte.so/workflows interactive
⌁ Prompt
“Alpine trail-running capsule, cold dawn light”
▣ Image input
✦ Generate image
Nano Banana 2.1 · 4:5
⌕ Generate video
Kling 3.0 · 5s
◎ Output
Waiting…
Nodes light up as the run reaches them — results render inline on the canvas.

Watching a run​

Each node shows its own state — waiting, running, done — and results appear directly on the node as they land: generated images and frames render inline, text steps show what they wrote. Downstream nodes start the moment their inputs exist, so parallel branches (say, twelve shots off one Split List) run side by side.

Need to bail out? Stop cancels the run; finished nodes keep their results, and you're only billed for work that actually executed.

Change one thing, pay for one thing​

Every step remembers what it was made from — its settings and everything wired into it. When you press Run again, a step whose recipe hasn't changed keeps its last result (it's marked Kept from your last run) and costs nothing; a step you edited runs again, along with everything it feeds.

Select any step for two more ways to run:

ActionWhat runs
Run this stepA fresh take of just this step, on the inputs it already has. Nothing downstream runs — check the new result first.
Run from hereA fresh take of this step and everything it feeds. Everything upstream is kept.

To start over, open the menu beside Run and choose Run everything again. Results only carry over while their files are still in your Library — delete one and that step simply runs again.

The Runs menu​

Click Runs in the top bar for the workflow's history. Every run is listed with a status:

StatusMeaning
RunningStill executing — select it to watch live.
CompleteEvery node finished.
FailedA node errored — its message is on the node; the rest of the graph kept whatever finished.
StoppedYou cancelled it mid-flight.

Selecting a run hydrates its per-node results back onto the canvas, so you can compare yesterday's outputs with today's, or retrieve a text step's exact wording from a week ago.

Where outputs land​

Everything a run generates is saved to your Library alongside your Studio generations — ready to drag into a cut, feed to an Application, or download. Wiring your finished results into Output nodes keeps the deliverables obvious in each run's results, which matters more as graphs grow.

Costs and failures​

A run bills each generator node as it executes — model choice sets the price, and the usage ledger itemizes every node's spend. If a run fails immediately, check your credit balance first; for other failures, see Generation issues.

Run it from anywhere​

Saved workflows are first-class API citizens: trigger them with custom inputs from scripts, backends, or agents like Claude Code — statuses and per-node results included. See Workflows via the API and MCP.