Run

Limits and credits

Each limit below is enforced by the engine. The last column says what happens when you reach it, because that is usually the question.

Run limits

Limit	Value	At the limit
Steps in one run	1000 positions. The trigger, every condition, switch, iterate, loop check, jump, wait and action each take one, and so do steps replayed after a wait. A Try takes none, but the steps inside it do.	The run fails at once with no retry.
Steps replayed when a run resumes	1000	The run fails with no retry.
Attempts after an engine error	3 (waits of 30 and 60 seconds)	The run ends as failed.
Laps of a Loop	1 to 500; a new loop starts at 100	The step fails, so the run fails (a Try around the loop can catch it).
Items of an Iterate step	500	The run fails before processing any item.
Jumps of a Jump step	1 to 500; a new one starts at 3	The run continues with the next step.
Chain of runs (signals, Trigger workflow, events)	10 hops	The firing or triggering step fails; an event is ignored.
One Wait step	30 days	Refused when you save.
One Wait until step	365 days ahead	The step fails when it runs.
Adjust date	3650 days either way	Refused when you save.
How long a worker may hold a run	5 minutes	After that the run can be taken over, and that costs an attempt.
Depth of a stored definition	100 levels	Refused when you save.

There is no maximum number of steps in a workflow: the limits are on what one run does.

Where	Limit	At the limit
Webhook body	256 KiB	413
Webhook rate	About 60 requests a minute per address (counted per server instance)	429 with `Retry-After`
Webhook Idempotency-Key	255 characters	400
HTTP step response	5 MB	The step fails
HTTP step time	15 seconds without activity	The step fails
HTTP step redirects	3	The step fails
Email search / calendar search	50 results	Truncated: the newest 50 emails, the 50 earliest events
Email inline attachments	About 2.7 MiB once encoded (about 2 MiB of files)	The step fails
Email from Filmo attachments	About 39 MiB encoded	The step fails
Run history shown	100 runs	Older runs are not reachable in the app
Runs kept	90 days after they end	Deleted with their steps
REST API / MCP	120 requests a minute per key	429
REST API / MCP idempotency	24 hours of replay	A repeat after that is a new run

Running a workflow uses the same pool of credits as the AI in Filmo.

One credit for every action step that completes, whatever it does. A wait, a Set variable and an Adjust date cost one, like an email does.
Free: the trigger, conditions, switches, loops, iterations, jumps, steps that fail, and anything replayed after a wait, a retry or a resume.
A step in the body of a loop or an iteration costs one on every lap.
A step that calls the AI costs that credit plus AI credits for the tokens it used.
On a retry, a step without AI is charged only once; a step with AI is charged on every attempt.
What you see are credits, never tokens: on the run card, in the run history and on each step.

Before each action step the engine checks the company's credit. The outcome depends on why it refuses:

Reason	Result	What you do
No credits left	The run is «Paused».	Add credit or turn on «Turn on pay-as-you-go», then press «Resume».
A payment problem on the subscription	The run is «Paused».	Fix the payment, then press «Resume».
Workflow execution is switched off for the company	The step fails, so the run fails.	Ask whoever administers the AI settings to turn it on again, then Retry.
No subscription found	The step fails.	Contact support.

The check asks whether any credit is left, not whether there is enough for this step, so the last step can take consumption slightly over the limit.
A pause costs nothing, writes no error and sends no notice. The run continues at the step that was blocked.
Plans with no ceiling never reach the limit. A monthly budget on pay-as-you-go still stops the run at the budget.
If the billing service itself is down, the step goes ahead rather than pausing every run of every company.

Run
Errors and retries
Two different things go wrong in a run: a step reports a problem, or the engine itself breaks. They are treated differently, and the difference is the most important thing on this page.
Build
Branches and loops
Three steps repeat work (and a fourth, Try, keeps a run going after a failure): Loop repeats while a condition holds, Iterate steps run their body once per item of a collection, and Jump goes back to an earlier step. Each has a limit, and each limit is enforced.
Run
Runs
A run is one execution of a workflow. Every run keeps what it did, step by step, so you can see where it stopped and why.

Was this page helpful?