Structured Data

As you chat with the AI on each node, it extracts key information and stores it as structured data. You never fill in a form — the fields populate from natural conversation.

How extraction works

The AI extracts structured data in the same pass as generating its conversational reply. When it detects information that maps to a known field for the current node type — a persona name, a tech stack choice, a feature priority — it writes that value to the node's structured data store simultaneously with streaming its response.

There is no separate extraction step. One AI call handles both.

Viewing extracted data

Open any node's side panel. Below the chat input you'll find the Extracted Data section. This expands to show all fields that have been populated for that node type.

Fields are grouped by category. For example, a Tech Stack node groups its data into Frontend, Backend, Database, and Infrastructure sections.

Empty fields are shown in a muted state — they indicate what information is still missing. You can direct the conversation toward those fields, or leave them empty if they are not applicable to your project.

Editing extracted data

You can edit any field directly in the Extracted Data panel. Click a field value to open an inline editor.

This protection is intentional. It lets you correct AI mistakes or provide information the AI cannot infer, without worrying that the next message will revert your edit.

To remove the protection and allow the AI to update a field again, clear its value in the panel. An empty field is always writable by the AI.

How extracted data feeds the PRD

Every field in the Extracted Data panels across all your nodes is available to the PRD generator. When you click Generate, the AI uses these structured fields — not a summary of the conversation — as the authoritative input for each PRD section.

This means the quality of your structured data directly determines the quality of your PRD. The more complete your nodes, the more detailed and accurate the generated document.

Merge behaviour

When the AI extracts new data, it merges it intelligently with existing values rather than replacing them. New information is added alongside what is already there. Fields you have manually edited are always protected — the AI will never overwrite your explicit choices.

Tips for richer extraction

• Be specific. "Our users are solo founders" extracts a persona. "We have users" does not.
• Answer follow-up questions. When the AI asks for clarification, those answers are where extraction happens.
• Mention constraints explicitly. Saying "we need this to work on mobile too" will populate mobile-related fields in your Tech Stack node.
• Connect nodes before chatting. The AI extracts more confidently when it has context from related nodes.