Skip to main content

Run a workflow

This page describes running a Nomic workflow or a saved workflow—not designing one. For the builder form and the AI design assistant, see Workflow Builder.

Start a run​

  1. Go to the Workflows tab in Nomic.
  2. Choose a workflow—Nomic workflows appear on the Nomic tab and alongside your saved workflows on All.
  3. Start the run. You may be prompted for requested files (per-run inputs defined in the workflow). Attach or select files from project context as instructed.
  4. Wait for the run to finish. Extra High workflows run until the task is complete; others wait between turns. You can often send follow-up prompts after the first pass to go deeper (especially for reviews).

Workflows respect the same permissions as the rest of Nomic: the agent only sees files you can access in the project or session.

Runners always get the live version. If you can edit the workflow, Run draft on the Versions tab lets you try unreleased edits on real inputs without changing what everyone else runs. The run page shows a Draft badge on those runs. See Workflow management.

Requested files vs context​

  • Requested files — Picked each time you run the workflow (this submittal, this drawing set). Described on the workflow card and in the run UI.
  • Context (configured in the builder) — Persistent files or folders attached to the workflow so every run sees the same reference material (for example, master specifications or firm standards).

Requested-file slots that accept multiple inputs can include folders as well as individual documents. Nomic expands each selected folder to its indexed document children at runtime. Single-file slots accept documents only.

If something is missing, the UI usually reflects what the workflow author asked for—good workflow descriptions on the card reduce back-and-forth.

Outputs and follow-ups​

Workflow runs produce artifacts—documents, tables, or drawing-review outputs—shown in the artifact panel. You can download (Excel, JSON, PDF markups where applicable) and often iterate: send another prompt to tighten scope, add checks, or resolve follow-ups without losing prior context.

Cost and duration

Long runs use more tokens and time. See Models & pricing for rough per-run expectations; scope projects and context to avoid pulling in unnecessary files.