Streaming First
State-machine incremental parsing — render as bytes arrive. The first character starts painting the DOM, no need to wait for the full response. A perfect match for token-by-token AI output.
The world's first For AI & zero-dependency streaming UI framework. Describe components with a minimal DSL on the backend, push over SSE or WebSocket, and render incrementally on the frontend — the first token starts painting the DOM, letting AI produce more flexible, expressive UI with minimal tokens.
On the left is the TokUI DSL source (auto-formatted + syntax highlighted); on the right is the real DOM it renders. Hit "Edit" to tweak it live.
LLMs emit plain text, leaving users staring at a wall of words. TokUI gives the model a structured expression language: stream rich, interactive UI from minimal tokens.
A state-machine incremental parser — a [card tt:…] tag creates its container on arrival, children fill in as they come. No waiting for close tags; first paint is instant.
Minimal DSL syntax (key:value, comma multi-values, boolean attrs) — the same UI costs far fewer tokens than HTML or a JSON schema.
Server TokUIBuilder chains the DSL → SSE push → frontend TokUI incremental render. One component vocabulary across the stack.
Built-in tool-call, thought chain, code diff, agent status, artifact preview and more — the interaction patterns of mainstream AI products, ready to go.
Every category ships full docs: prop tables, variant notes, and editable live examples.
Import the build, mount a container, feed data — that simple.