Duplicating, exporting and importing flows
Copy a flow to experiment safely, or move one between workspaces as a JSON file — with keys, pool bindings and webhook URLs deliberately left behind.
In this article (4)
Three separate things people often conflate: duplicate (a copy here), export/import (a file you can move), and share (a read-only view for someone else). This article covers the first two.
Duplicate
The copy icon on a workflow card makes an independent copy in the same workspace. Nothing is linked afterwards — edit the copy freely. The obvious use: try a different model or a rewritten prompt side by side with the version people rely on.
Export
The download icon gives you the flow as a JSON file: its cards, wiring, instructions, models and settings.
What is deliberately not in it:
- No API keys and no credentials — ever.
- No pool bindings you could use elsewhere; the importing workspace picks its own pools.
- No webhook URLs from Send-to-a-URL cards, because those URLs usually are a secret.
- No run history, results or items.
So an export is safe to email, commit or hand to another workspace, and it is a reasonable belt-and-braces backup of your own work.
Import
Workflows → Import, then choose a .json file. It arrives as a new flow that you own.
Before it can run for real, open it and set:
- A pool on each AI step (the export carried none).
- Any webhook URL on a Send-to-a-URL card.
- A connected server and tool on any Call a tool step — see MCP servers.
Until then the flow is complete but unpointed — it will not spend anything by accident.
Which to reach for
| You want to… | Use |
|---|---|
| Try a variation without risking the original | Duplicate |
| Move a flow to another workspace or machine | Export, then Import |
| Let a client or colleague see a flow and its results | Share |
| Run the same flow from a script or another system | A trigger or the API |