Documentation
Publishing and versions: draft vs published workflows
Every workflow has a draft you edit and, once you publish, a version that the outside world runs. Keeping the two apart means you can change prompts and rewire nodes without breaking a schedule or an app that calls the workflow.
The draft
The canvas autosaves as you work, and what it saves is the draft. Run in the editor and Test this node always use the draft, so you can try a change the moment you make it.
Publish a version
Press Publish in the editor toolbar. Worfilo validates the draft the same way a run does; anything that would stop it running blocks the publish, while warnings, such as a tool list that may be out of date, are shown but do not.
A successful publish freezes the draft into a numbered version ("Published version 3") and makes it the active version straight away. A version never changes after it is made: later edits only ever change the draft, until you publish again.
What runs which version
- Editor Run and Test this node: the draft.
- Playground: the published version by default, or the draft if you switch it. A workflow that has never been published can only use the draft.
- Triggers: always the published version. See Schedules and triggers.
- Workflow API: always the published version. Calling a workflow that has never been published returns
409. See Workflow API.
Going back to an earlier version
Earlier versions are kept. To go back, the simplest way is to undo the change on the canvas and publish again, which makes a new version with the old behaviour.
A coding agent connected through the Worfilo MCP server can also activate an earlier version directly with its activate_version tool, which makes that version the one triggers and the API run.
Runs remember what they ran
Each run stores a snapshot of the graph it executed. Replaying an old run shows that graph, not today's draft or version, so you can always tell what a past run actually did. See Runs and replay.
Keep reading
- Workflow API: run workflows from your own codeCall a published workflow over HTTP with an API key and its workflow ID, send it input and get its output back, and test it first in the chat playground.
- Schedules and triggers: run workflows automaticallyRun AI agent workflows on a schedule: pick daily, weekly or a cron expression, set a timezone and payload, preview the next runs, and run it now.
- Runs and replay: watch AI workflows run step by stepRun an AI agent workflow and watch each node's status live, inspect its inputs and outputs, test one node, and replay any run on the graph it used.
Build it on the canvas
Create a free account, describe the workflow or wire it yourself, and run it in the browser.