Point the desktop app at a real codebase and kratai generates a spec for it automatically.
Open a project
Sign in to the desktop app, then open the folder of a real codebase — a new project or an existing one, it doesn't matter which. kratai reads the project locally; nothing about this step sends your code anywhere.
What happens on first open
The first time you open a project, kratai automatically:
- Generates a Use Case Model — actors, use cases, and their relationships — from your routes and folder structure.
- Generates a Data Model — entities and relationships — from your schema and data-access code.
- Auto-chains straight into a detail-fill pass: chat reads the actual files behind each use case and fills in its goal, preconditions, main flow, and postconditions, rather than leaving them as a one-line guess.
Each step is a real AI call and draws on your account's credit balance — see Installation.
Where to look
Once generation finishes, you have a few views to explore:
- Spec — the dropdown that switches between the Use Case Model, Data Model, and SRS document views below.
- Use Case Model — the actor/use-case diagram, with each use case's detail available on click.
- Data Model — the entity-relationship diagram.
- SRS document — the same spec rendered as a formatted document, exportable to PDF.
- Class Diagram — generated separately by static analysis, no AI involved, always reflecting the real code.
Ask it something
Open chat and ask about a specific use case or entity — kratai reads the actual files behind it before answering, rather than guessing from the diagram alone. This is also how you edit the spec going forward: see Manage Specs.
Next: using kratai with your AI coding agent
You have a real, grounded spec sitting in your project now. See How kratai Works for how it connects to whatever AI coding agent you already use to write code.