Dendrite SDKIn-Browser AI Chat

Run LLM personas entirely in the browser. WebGPU where it works, WASM where it doesn't.

Zero Server (by default)

Models run directly in the visitor's browser via WebGPU or WASM. No API keys, no backend, no data leaves the device.

Picks the fastest engine available — WebLLM on WebGPU, Transformers.js on WASM, or an optional cloud proxy if you want one.

Shape LLM behavior with typed documents — zero-shot identity, personality, knowledge, and instruction docs.

Pure TypeScript core. Drop-in Vue components in @agent-layer-zero/dendrite-vue, React in @agent-layer-zero/dendrite-react.

Async generator API streams tokens as they're generated. Build responsive chat UIs with real-time output.

Looking for prebuilt UI? See the Framework Packages guide for dendrite-vue, dendrite-react, and soma (shared CSS).