Documentation
The builder
Plain language in, a complete workflow out. It is the only thing that edits a workflow.
How you use it
You write what should happen. For example: onboard new corporate clients, read the KYC pack, screen it against our compliance rules, get sign-off on anything high risk, then file the dossier and notify the team. The builder asks for whatever it still needs, then builds.
It works in a conversation attached to the workflow, so the reasoning behind every change stays with the thing it changed.
What it authors
- Agents, with their instruction, their allowed commands and their memory.
- Steps that call a command, and the conditions that choose between branches.
- Approval gates, and who they go to.
- Triggers that start the workflow.
- Error handling regions, loops, and the response a webhook sends back.
- Notes on the canvas, for the people who will operate it.
What it can see
To build well the builder can look at your knowledge while authoring: it can list knowledge bases and documents, read a document, and search. That is how it knows what your material actually contains.
- It cannot read a knowledge base marked sensitive, at any point during authoring.
- It never sees a credential. Test runs execute without them.
- It cannot reach your connected systems while authoring.
Why only the builder edits
An operator who can hand-edit a live workflow is a control gap: changes arrive with no request behind them and no record of why. In PromptOne the operator seat is read-only by design, so every change is described, made in one place, and recorded.