Design system — the same screen, rendered in each
0s
0
0
0
–
· –
Point Catalystic UI at your own endpoint — it gets the conversation and returns UI in your selected design system. Any framework works.
POST { history, components, mode }{ reply, components } — an A2UI tree, or [] for a text only replyhttps://api.dify.ai/v1/chat-messagesImplement a POST endpoint that speaks the contract — pick your language.
Drop this into your own frontend. Your agent returns an A2UI tree; Catalystic UI renders it as real, interactive UI — taps flow back to your agent.
Give your agent a name and describe what it's for — Catalystic UI uses this to shape how it responds and what it builds.
Tweak how this agent behaves. Saved in this browser; applies from your next message.
Keys authenticate calls to the generative-UI API. Shown once at creation, hashed at rest — revoke anytime.
Every API call is metered against your balance — about £0.01 per fast generation, £0.05 for quality.
Call the same generative-UI engine that powers this playground from your own backend.
POST /api/chatAuthorization: Bearer ck_…{ "prompt": "…", "mode": "chat" }{ reply, components, credits_remaining }Keep a record of the key below. You won't be able to view it again.
Found a bug, or have an idea? It goes straight to the team.