Architecture modelling for spec-driven development
Draw your app. Generate a spec your AI can build from.
One canvas for tables, screens and flows — saved as plain text that diffs like code. Paste your schema and see it.
FREE edition: 2 diagrams, UNLIMITED offline diagrams — every import and export!
UAML Studio is different from what i have seen, it is easy to use and pretty powerfull out of the box — Patrick W..
Diffs like code
Your whole app's architecture in one plain-text file that lives in your repo and reviews in a pull request. No binary board, no docs that go stale.
Convert anything
Paste your schemas — PostgreSQL, MySQL, MariaDB, SQL Server, Oracle, SQLite, DuckDB, Prisma, Drizzle, DBML, Mermaid, BPMN 2.0, PlantUML — and convert to UAML.
Feed your agent a spec, not a prompt
One click — generates a deterministic specification with real architecture artifacts. Review it, accept it — stop explaining your system every session. One-shot your apps with accuracy.
How it works
Draw it. Generate the spec. Hand it to your AI.
Three steps, and the middle one is the product. What comes out is a specification your agent builds from — not an application.
- 01
Draw your app
One canvas holds tables, entities, screens, forms, flows and integrations. Eight node kinds and five wire families — a closed alphabet, so there is nothing else to learn. Or paste an existing schema and start from what you already run.
- 02
Generate the spec
The generator prints a deterministic Markdown constraint set: entities, fields, relations, screens and flows, chaptered by your namespace tree. You review it and accept it before anything is written.
- 03
Hand it to your AI
Drop the specification into Claude Code, Cursor, Copilot or whatever you build with. The structure and the constraints come from your model; the design, the copy and the algorithms stay yours and your agent's. It is a specification, not a build button.
What is in the box
Start from your real system
Everything below runs today unless it is marked
Import from SQL
PostgreSQL, MySQL, MariaDB, SQL Server, Oracle, SQLite, CockroachDB and DuckDB. Paste a dump; the importer works out the dialect.
Import from ORMs and DBML
Prisma and Drizzle schemas, plus DBML — the formats your repo already carries.
Import from Mermaid
All eight Mermaid diagram types — ER, class, flowchart, state, journey, timeline, C4 and architecture — become editable models, not pictures.
Import BPMN 2.0 and PlantUML
Process models and UML from the tools your organisation already standardised on.
Deterministic exports
.uaml, the Markdown specification, SQL per dialect, DBML, Mermaid, PlantUML and images. The same model exports byte-identically every time, so a diff means a real change.
Share links
Private, invite-only or public links to a cloud-saved diagram. Read-only for the reader, revocable by you.
The AI agent, your key or ours
Bring your own Anthropic key and pay nothing extra, or spend credits and skip the key management. Both run the same agent against your model.
Skill-pack export
A pack that drops into .claude/skills/ and keeps a coding agent inside your architecture for the whole build.
Pricing
Four rungs, and the bottom one is free
Prices in EUR, excluding VAT. Everything modelling-related is in the Free edition; you pay for cloud, scale and seats.
AI credits are optional and start at €5 — or bring your own Anthropic key and pay us nothing for the agent. See the full ladder →
Draw the system you are about to build.
Paste a schema, hit auto-arrange, generate the spec. It takes a few minutes to find out whether this fits how you work.