Appearance
Plan & Implement
This guide explains how to use the project assistant to design and build the workflows that deliver a project, and how to hand the same work to Claude Code instead when you'd rather work there.
Overview
The Plan & implement tab is a conversation with an assistant that works inside your environment. It reads the project — its connectors, reference documents, and (for a migration) its profile and data quality report — and it can create and change workflows as you agree them.
As it works you see what it is doing: each tool it calls appears as its own line, with a tick when it succeeds and a cross when it doesn't, and a running token count sits under the conversation.
The assistant acts as you
It can only do what your own permissions allow, and writing anything requires agent write access on the environment. A failed tool call tells you which of those stopped it.
Work with the assistant
- Open the project and select the Plan & implement tab
- Type what you want, or pick one of the suggestions
- Read what it proposes and reply — it works in a loop, not a single pass
The assistant is asked to propose a design and wait for your confirmation before building, to prove the design on a small slice before building the rest, and to tell you plainly what it has verified and what it has only assumed. Hold it to that: a tool reporting success is not the same as the result being right.
Your conversation is your own — another person on the same project sees theirs, not yours. Use the + button to start over and the clock button to return to an earlier conversation.
Build it in Claude instead
Some work is easier in Claude Code, where the repository and your own tools are in reach. Copy prompt for Claude copies a briefing for the project — its goal, its identity, and the working rules above — ready to paste into a new conversation in Claude Code or Claude Desktop.
Claude needs access to your platform
The briefing tells Claude to read the project from YuzeData, so the Claude you paste it into must be connected to your YuzeData environment. Without that connection it can't reach the project's connectors, documents, or profile.
Fill the project in first
Both routes are only as good as the context the project holds:
| Fill in | Why it helps | Where |
|---|---|---|
| A specific Description | It's the assistant's statement of the goal — the clearer it is, the better the plan | Overview tab |
| Linked connectors (with roles) | Shows which systems are in play | Link Connectors and Workflows |
| Reference documents | Data dictionaries and requirements give the business context the data alone can't | Add Reference Documents |
| A completed profile and data quality report | Says what the source actually looks like and what needs attention (Migration) | Profile a Database |
A thin project produces a thin plan
A project with a vague description and no linked systems gives the assistant almost nothing to work from. One with a clear goal, its connectors, and its reference documents lets it start on the actual work.
Related guides
- Projects — what a project is and what each tab does
- Link Connectors and Workflows — connect the systems the project delivers
- Add Reference Documents — give the assistant business context to work from