assistant-ui is an open-source TypeScript/React library that provides composable primitives for building production AI chat user interfaces. It offers a CLI for scaffolding projects and supports rendering tool calls as React components with inline human approvals.
Project overview
It delivers composable UI primitives rather than a single monolithic chat component, enabling developers to build AI chat experiences in React while supporting generative UI patterns such as rendering tool calls as React components and collecting inline human approvals.
Use cases
Coding & Development
Deployment
Refer to project documentation
License
MIT
Best for
Developers with TypeScript and React knowledge who need composable primitives to build AI chat interfaces.
Teams that need generative UI support such as rendering tool calls as React components and collecting inline human approvals.
Key capabilities
Render tool calls and JSON as React components, collect inline human approvals, and expose safe frontend actions to the model.
CLI tool to scaffold a Next.js app or add styled components to an existing project.
Limitations and risks
React and TypeScript coding knowledge is required to use the library.
An AI model backend such as OpenAI, Anthropic, Google Gemini, or Mistral is required; the library does not include its own model.
Setup involves medium difficulty due to the need for TypeScript and React knowledge, CLI-based scaffolding, and configuration of backend and runtime adapters.
Getting started
Run npx assistant-ui@latest create to scaffold a new project, configure a backend runtime such as useChatRuntime for Vercel AI SDK, and import and render the Thread component.
Setup requires TypeScript and React knowledge, familiarity with the CLI scaffolding workflow, and configuration of backend and runtime adapters.