EN 日本語
Kurage

Kurage Architect

From conversation to implementation-ready design docs.

"I know what I want to build, but the requirements won't come together." Kurage Architect grows that vague idea into a system design document you can implement from directly — by talking it through with AI. Open source, powered by a local LLM (Gemma 4).

Sign in with X. No credit card required.

FEATURES

What it does

💬AI elicits your requirements

Asks 1–3 focused questions at a time, separating decisions, proposals, assumptions and open questions.

🗺️Diagrams generated for you

Architecture, class and sequence diagrams in Mermaid, embedded straight into the design doc.

📄Hand-off-ready output

Export as Markdown, requirements JSON, HTML or PDF — a perfect input for AI coding sessions.

🔒Local LLM, private by design

The brain is a local Gemma 4. Your ideas are never sent to external AI providers.

🧭Spec-driven flow

Your design matures through 7 stages: discover → clarify → specify → plan → design → review → ready.

🛟Your answers are never lost

Even if the LLM fails, your answers are saved. Resume the design when it recovers.

HOW IT WORKS

Three steps

Sign in with X

No registration. Design projects are isolated per X account.

Describe what you want

A short description is enough. Kurage asks the right questions to firm up requirements.

Receive the design doc

Download as Markdown / JSON / Mermaid / PDF and go straight to implementation.

PRICING

Pricing — the first one is free

YOUR FIRST DESIGN
Free
first project
  • ✅ Full feature set (conversation, diagrams, export)
  • ✅ First document export (MD/PDF/JSON/Mermaid) is also free
  • ✅ No credit card required
  • ✅ Start now with just an X login
Payments use the same rails as the Kurage blog's paid articles (server-side PayPal order verification; on-chain verification on Base for URLAI). URLAI is a project token with a growing list of places to use it — you can also earn it for free via kurl2earn.
🪼 Start free (first project free) Open to everyone. Your designs are processed by a local LLM and never sent to external AI providers.