Run an action

Launch a single action over a space’s files — one file or many — and follow it on the run page.

View as Markdown

Once an action is installed and configured (see Actions and the library), you run it over the files in a space. A run reads the files you pick, does the action’s one job, and shows results on a run page as they come in.

For structured extraction specifically, see Extract data; for filling and signing, see Document generation.

Where runs start

Every standalone run lives in a space’s Actions tab. Open a space and select Actions — it lists “Every action run in this space.” with a “Run action” button. If there’s nothing there yet you’ll see “No runs yet” / “Run an action over this space’s files to see results here.”

You can also launch a run from the Files tab. Select one or more files, and the bulk toolbar’s “Run action” button starts a run over exactly those files.

The launcher

“Run action” opens the launcher, titled “Run an action” (“Pick an action and the files to process. Results appear on the run page as they come in.”).

1

Pick an action

Use the “Action” select (“Choose an action”). If nothing is installed, you’ll be pointed to “the actions library.”

2

Fill in run settings

If the action takes per-run settings, a “Run settings” form appears. Fill it in.

3

Choose files

Pick the files to process — the picker shows {n} of {m} selected” with “Select all eligible files” and “Upload more”. Files the action can’t handle show “This action can’t process this file” and can’t be selected. An action that works on one file at a time accepts exactly one.

4

Run it

Click “Run action”. You’re taken to the run page, where results fill in as they arrive.

The launcher also shows a “Recent runs” list so you can jump back into a run you started earlier.

The Run an action launcher with the 'Invoice extraction' action chosen, its Run settings form, and 2 of 5 files selected

The run page

The run page (/spaces/$spaceId/runs/$runId) headers the action name and a status badge — “Running”, “Waiting for you”, “Done”, or “Failed” — alongside the start and finish times and a source-file count. While a run is active you can “Cancel run” (confirm: “Cancel this run? The run stops and is marked failed. This can’t be undone.”). If the AI assistant dispatched the run from a playbook, a “Triggered by playbook run” link points back to it.

The body of the page changes to fit the action:

ActionWhat you see
Structured extraction (running)Live per-file progress — {done} of {total} documents extracted”, each file flipping from “Extracting…” to “Extracted”.
Structured extraction (done)The results grid — see Extract data.
Fill form (paused)The in-app “Review form values” step — see Document generation.
SignThe “Signatures” envelope panel — see Document generation.
Everything elseGeneric sections: “Documents” (the source files), “Results”, and “Generated files” (download rows).

For redaction and other file-producing actions, the redacted or generated file appears under “Generated files” — download it from there.

An action run page in the Done state, showing the run header with start and finish times and the Results section below

Bulk runs

To process many files at once, open the space’s Files tab, select the files, and use the bulk toolbar’s “Run action”. This launches a single run over every selected file — the same run page, with one row per file. There’s no separate “batch” mode; a bulk run is just a run with many source files.

The raw output, the list of generated files, and the action version are folded into a developer-mode-only disclosure labeled “Run details — sources, generated files, raw output.” Turn on developer mode from the profile menu to see it. See Actions and the library.

Where to go next