Kommit Docs

Generating a PRD

How to generate a production-ready Product Requirements Document from your canvas nodes.

Generating a PRD

Once you've had conversations on your key nodes, Kommit can assemble everything into a structured Product Requirements Document. The PRD is generated by AI but grounded entirely in the structured data you've built up on the canvas.

How to generate

  1. Open any node's side panel.
  2. Click the PRD tab at the top of the panel.
  3. Click Generate PRD.
  4. The document streams in real time — you can read it as it forms.

Generation typically takes 20–40 seconds for a typical project. Larger canvases with more nodes and richer conversations produce longer, more detailed PRDs.

What goes into a PRD

The generator assembles:

  • Project node — executive summary, problem statement, goals
  • Audience node(s) — target user profiles and success metrics
  • Competitors node(s) — competitive landscape and differentiation
  • Tech Stack node — architecture decisions and infrastructure plan
  • Feature nodes — full requirements per feature, with user stories, acceptance criteria, and complexity
  • Assets node(s) — design references and annotated assets

Each section of the PRD maps to a node type. Empty nodes produce thinner sections; fully-populated nodes produce detailed, specific content.

Repository context

If you've connected a GitHub repository to your project, a toggle appears before generation: Include repository context.

When enabled, the AI incorporates your codebase into the PRD — referencing existing patterns, current database schema, and established conventions. This is especially useful when you're adding features to an existing product rather than starting from scratch.

Enable repository context when the PRD is intended for an agent that will work directly in your existing codebase. Disable it for greenfield projects to avoid confusing the AI with unrelated code.

PRD structure

Generated PRDs follow a consistent section structure:

  1. Executive Summary
  2. Problem Statement
  3. Target Users
  4. Competitive Landscape
  5. Technical Architecture
  6. Features & Requirements
  7. Acceptance Criteria
  8. Design & Assets
  9. Open Questions

The AI writes each section in prose, using the structured data from your nodes as the authoritative source of facts.

Version history

Every generation is saved automatically. To view previous versions:

  1. Open the PRD tab.
  2. Click History.
  3. Browse the version list — each entry shows a timestamp and the number of nodes included.
  4. Click any version to read it in full.
  5. Click Back to return to your current PRD.

Version history does not have a fixed limit per project. All previous generations are retained. You can compare versions by opening them side by side in different browser tabs.

Regenerating

Click Regenerate to produce a new version at any time. This is useful when:

  • You've had additional conversations since the last generation
  • You've added or removed nodes
  • You want to try with repository context enabled or disabled
  • You're iterating on the spec after stakeholder feedback

Each regeneration creates a new version in history. Previous versions are never overwritten.

Aborting generation

If a generation is in progress, an Abort button appears in the panel. Clicking it stops the stream. The partial output is discarded and you are returned to the Generate view. No version entry is created for aborted generations.

On this page