assistant-ui is an open-source TypeScript and React library, released under the MIT licence, for building production-grade AI chat experiences quickly, and it is aimed at frontend and full-stack developers who need the user experience of ChatGPT inside their own React application.

What it is

assistant-ui lives in the React and TypeScript ecosystem and ships as composable primitives rather than a finished product. The library provides Thread, Message, Composer, ThreadList, and ActionBar components, and its CLI copies a styled shadcn/ui theme into the project so every pixel stays under the developer's control. Backed by Y Combinator, it is distributed on npm as @assistant-ui/react alongside backend adapters such as @assistant-ui/ai-sdk.

The concrete problem it solves is the monolithic chat component. Instead of dropping in one opaque widget that cannot be restyled or extended, developers compose primitives and supply their own styles, runtimes, and backends. useChatRuntime connects to the Vercel AI SDK out of the box, while useLangGraphRuntime, useDataStreamRuntime, or a custom runtime can be swapped in for an existing backend. This is the piece of UI that would otherwise be written from scratch for every project.

Key capabilities

• Composable primitives — Thread, Message, Composer, ThreadList, ActionBar — assembled into any chat layout instead of configured through props.
• Production interaction behaviour out of the box: streaming, auto-scroll, retries, attachments, markdown, code highlighting, voice dictation, keyboard shortcuts, and accessibility.
• Generative UI that renders tool calls and JSON as React components, collects inline human approvals, and exposes safe frontend actions to the model.
• End-to-end TypeScript typing across runtime APIs, tool schemas, message parts, and adapters.
• Backend adapters for the Vercel AI SDK, LangGraph and LangChain, the AG-UI and A2A protocols, Google ADK, OpenCode, and custom data-stream backends.
• Model support for OpenAI, Anthropic, Google Gemini, Mistral, Perplexity, AWS Bedrock, Azure, Fireworks, and Ollama, plus community providers through the AI SDK.
• assistant-cloud for managed thread history, telemetry, and file storage.

Who uses it and how

• Teams placing an AI copilot or chatbot inside an existing React application where the chat surface must match an established design system; the shadcn/ui theme is copied into the repository, so styling stays local.
• Products already running an agent backend on LangGraph or LangChain that need a typed React front end for it.
• Applications requiring inline human approval, where generative UI renders a tool call as a component and the user accepts or rejects it before execution.
• Projects that route model traffic through the AI SDK to OpenAI, Anthropic, Gemini, Bedrock, Azure, or a local Ollama instance to keep model choice open.

Getting started

The fastest path is the CLI: npx assistant-ui@latest create scaffolds a new Next.js project, and npx assistant-ui@latest init adds the styled components to an existing one. Direct installation uses npm install @assistant-ui/react @assistant-ui/ai-sdk, after which AssistantRuntimeProvider with useChatRuntime and the generated thread.aui element renders a working thread.

How it compares

No list of paid products that this project replaces is provided in this entry, and no comparable library is named alongside it, so it stands alone in this registry. Evaluation therefore has to run against whatever alternatives a team is already considering.

When to use it — and when not to

Choose assistant-ui when the chat interface is a first-class part of a React product, the visual design must be owned rather than themed from outside, and the backend runtime is a deliberate architectural choice. Self-hosters must supply that runtime themselves and wire the HTTP backend, and managed thread history, telemetry, and file storage require adopting assistant-cloud. Teams working outside React and TypeScript should look elsewhere, and 207 open issues signal a large surface to track; the production-user list in the README is blank in this entry, so that claim cannot be checked here.

npx assistant-ui@latest create   # new project
npx assistant-ui@latest init     # add to existing project

npm install @assistant-ui/react @assistant-ui/ai-sdk

"use client";

import { AssistantRuntimeProvider } from "@assistant-ui/react";
import { useChatRuntime } from "@assistant-ui/ai-sdk";
import { Thread } from "@/components/assistant-ui/elements/thread.aui";

export function Chat() {
  const runtime = useChatRuntime();
  return (
    <AssistantRuntimeProvider runtime={runtime}>
      <Thread />
    </AssistantRuntimeProvider>
  );
}