Running

Run a flow or a single neuron, follow it while it runs, stop it, and read the run log.

Running a workflow means starting one of its neurons: that neuron runs, and every neuron connected after it starts on its own as each one finishes. You can run from the canvas's top bar, from the Neurons section, from a neuron's START tab, or from the Workflows page without opening the canvas.

Run from the top bar

The run button sits at the right of the canvas's top bar. What it runs depends on what is selected:

With one flow, it reads Run flow and runs that flow from its first neuron.
With several flows, it reads Run plus the name of the selected neuron's flow. With nothing selected, it runs the first flow.
With a neuron selected that isn't connected to any other, it reads Run plus that neuron's name.

When the workflow has several flows, open the arrow beside the button (Choose a flow to run). Each flow is listed with how many neurons it has. Pick one to run it, or choose Run all flows to start every flow at once.

On a narrow window the button just reads Run.

Run from the Neurons section

Open Neurons in the canvas's left rail.
To run a whole flow, click the play button on the flow's header (Run plus the flow's name).
To run one neuron, open its menu (Run and options) and choose:
- Play: runs the neuron and every neuron connected after it.
- Test this neuron only (shown when other neurons come after it): runs just this neuron. Nothing is sent, nothing is saved to memory, and the neurons after it don't start.
- Play all queued (shown when an input has queued values): runs the neuron once for each queued value, one after another.

A neuron that isn't in a flow has its own play button on its row. The same choices are in the neuron's START tab; see Neuron parts.

Before a run starts

When you press Play, the canvas first saves your changes, then checks the neuron:

It needs a value in at least one GET input. Instructions in DO alone are not enough.
No connection on the way into it may have a broken link (see "Broken links" below).

If a check fails, the neuron's result shows why and nothing is billed.

When a run asks for an answer

An input set to Ask when run asks for its value. When you press Play on the neuron, you're asked first, and Cancel runs nothing. When the neuron comes later in a flow, or starts from a schedule, an app event or a webhook, the run waits at that neuron: the neuron shows Waiting for you, the status card at the bottom reads Waiting for you with Answer and Stop (no timer runs), a notice says "needs an answer to go on" with Answer, and Runs lists it under WAITING FOR YOU. Click Answer and then Answer and continue to carry on, or Stop run to cancel it. The run waits for the time set in Wait for an answer, from 5 minutes to 24 hours; after that, the input's default value is used, or the neuron stops with "No answer given" and its On error link runs. See Neuron parts.

One run at a time

Only one run can go at a time in a workflow. While a run is starting, the top-bar button reads Running…; once a neuron is working, it turns into Stop. Meanwhile every play button is greyed out. Hover over it to see "Wait for the current run to finish".

Follow a run

While a run is going, a status card at the bottom of the canvas shows which neuron is working, for example "Running Summarise…", how long the run has taken, and "Calling …" when the neuron is using a connected app. Click Details to open the Runs section.

When the run ends, the card sums it up: "Nothing running in …" with how many neurons ran, how many had an error, the credits recorded and the time it took. After Test this neuron only, it reads "Test run · …" and "Nothing sent, saved or run after it". Click View usage to see what the run cost.

In the Neurons section, each neuron's row shows where it is in the run: Running while it works, Waiting if it comes after a neuron that is running or has finished and hasn't started yet, and Completed once it has finished in this run. Waiting goes away when the run stops or when nothing starts for a short while, and Completed goes back to the neuron's usual status a minute after the run ends.

Stop a run

Stop ends a run early. Only the Builder who owns the workflow can stop a run.

While a run is going, click Stop in the top bar, on the status card, on the running neuron's row or flow header in the Neurons section, or Stop run in the Runs section.
The status card reads "Stopping · … is finishing". A neuron that is already working finishes its current step; no new neuron starts after it, a running Code neuron ends in error with "Stopped." (no On error neuron starts, and the sandbox time it used is charged), and any values still waiting in a queue are dropped.
The card then reads "Stopped · …" with how many neurons ran before the stop.

Queues

A queue lines up several values for one input, and each value becomes its own run. Set it up on a Set in advance input in GET, then press Play all queued in the neuron's menu or Play through the queue in START. The Neurons section shows how many values are queued on each neuron. See Neuron parts for how to add values to a queue. A Code neuron can't play through a queue: loop over the items in your code instead (see Code). A single Play on it uses the current input.

Circular flows

A circular flow is a flow whose connections loop back, so its neurons can run again and again. Use one on purpose when a flow should repeat, for example to refine a draft until it is good enough.

When a workflow has a loop, the Neurons section shows a Circular flow card: "Connections form a loop, so this flow may run repeatedly. Monitor credit usage."

Loop shows which loop the current run is on, out of the most allowed.
Stop after sets how many loops a run may make before it stops by itself. Type a number and click away to save it; a number outside the allowed range is changed to the nearest one allowed. Hover the box to see the range. A number above the default shows a warning, because more loops can mean more credits.
Force stop appears while the loop is running. Click it to stop the run.

A circular workflow has no flow-level run: flow headers have no play button, and the top-bar button needs a neuron. Select the neuron the loop should start from, then click Run plus its name.

On the Workflows page, a workflow with a loop shows a Circular Flow badge, and its flow card shows Open Workflow to Manage instead of a run button.

Run from the Workflows page

You can run a flow without opening the canvas.

Open Workflows. Each workflow card shows its flows, one at a time (Flow 1, Flow 2, …).
Under Get, fill in or upload the flow's first input.
Click the play button (Run flow) on the flow.
The result appears under Make, with its Cost and Time.

Flows with a loop can't be run from the card.

Broken links

A broken link is a link on a connection that no longer fits, for example because the parent no longer has the item it passed, or the child's input changed type. A broken link stops the neuron it leads into and every neuron after it from running.

When a link breaks, a notice says "Broken link — Parent → Child: …. Open the connection to fix it."
The neuron after it shows Blocked in the Neurons section, and the connection line is marked.
Pressing Play shows "Fix the broken link before running — …".

To fix it, click the connection line to open the connections window, then remove the link or link the right item. See Connections.

The run log

The Runs section in the canvas's left rail shows what ran.

Open Runs in the left rail. The button at the top runs or stops the workflow, like the top-bar button.
LATEST RUNS lists the runs you started since you opened the workflow, newest first, by the neuron you played, with how long it took and whether it finished, failed or was stopped. A failed run shows its error. A Code neuron's run also shows the credits it cost and, for you only, Logs with what the code wrote. Click Logs to open them.
Each run shows the memories it used. Click one to open it in Memory; a memory that is no longer there opens the Timeline instead. Lines under the run say what it saved or proposed to memory, or why something wasn't remembered, and "Saving to memory…" while it is still saving.
A run that failed on a broken link or a broken reference has a Fix button that takes you to the setting.
USAGE FOR THIS WORKFLOW lists the billed operations from this workflow's neurons, with their credits and, for AI work, the model each one used. Test runs are marked "test run". Click See all usage to open the Usage page.

The run list starts empty each time you open the workflow. Usage stays.

Limits

Loops before a circular flow stops: 3 loops by default. You can set from 3 loops up to 50 loops.
Inputs per neuron: up to 3 inputs.
File upload: up to 32 MB per file.

Common problems

Play is greyed out

Something in the workflow is already running, and only one run can go at a time. Wait for it to finish, or stop it.

The run button says "Select a neuron to run"

There is nothing to run yet, or the workflow has a loop. Click a neuron on the canvas, then click the run button.

This neuron is busy with another run

The neuron is still working on an earlier run, for example one started by a schedule or a webhook. Try again in a moment.

Please provide at least one input with content and instructions

The neuron has nothing to work on. Give one of its GET inputs a value and write instructions in DO.

No input content provided

The neuron has instructions but none of its GET inputs has a value. Type or upload a value, then run it again.

Fix the broken link before running

A connection leading into this neuron has a broken link. Open the connection named in the message and fix or remove the link.

Could not stop the run

The stop request didn't go through and the run is still going. Click Stop again.

A circular flow stopped by itself

The run reached the number of loops set in Stop after. Raise the number if the flow needs more loops, keeping an eye on the credits each loop uses.

Nothing runs while I look at an older version

While you view an older version from Versions, nothing can run. Click Back to current first. See Versions.