Getting Started
Create a new OpenUI application or add OpenUI to an existing agent.
Start a new agent application with the OpenUI CLI, or add OpenUI to an application you already have.
Start a new application
You need Node.js 20 or later. Run the interactive setup to choose your model connection and agent framework:
npx @openuidev/cli@latest createThe CLI creates the project, installs its dependencies, and can start the development server for you.
Choose a model connection
Use OpenUI Gateway for model access, output correction, and provider fallbacks:
npx @openuidev/cli@latest create --name my-agent --template openui-cloudOr connect the generated application directly to your own model provider:
npx @openuidev/cli@latest create --name my-agent --template openui-self-hostedChoose an agent framework
Add --backend-framework to scaffold an integration for a specific agent framework. These options can be combined with either model setup.
npx @openuidev/cli@latest create \
--name my-agent \
--template openui-cloud \
--backend-framework defaultThe generated application includes Agent Interface and an OpenUI Lang component library. Follow the Agent Interface Quickstart to understand the generated project.
Add OpenUI to an existing application
assistant-ui or CopilotKit
Keep your existing conversation experience and add OpenUI Lang rendering.
Custom chat UI
Render generated interfaces inside your existing message list.
Agent backend
Route model requests through OpenUI Gateway and generate validated OpenUI Lang.
Build only with OpenUI Lang
Use OpenUI Lang without adopting a complete agent interface or hosted model service:
- Define the components the model can use.
- Generate the system prompt from the component library.
- Send those instructions to your chosen model or agent framework.
- Render the streamed response with your application components.
Learn more in the OpenUI Lang documentation.