The Open Standard for Generative UI
---
title: "assistant-ui"
---
<Steps>
<Step title="Register the OpenUI toolkit" source="https://www.openui.com/docs/api-reference/assistant-ui">
Inside AssistantRuntimeProvider, add openuiIntegration.toolkit through AuiConfig and pass openuiIntegration.instructions to useAssistantInstructions.
```tsx title="src/components/openui-config-provider.tsx"
"use client";
import type { ReactNode } from "react";
import {
AuiConfig,
AuiProvider,
Tools,
useAui,
useAssistantInstructions,
} from "@assistant-ui/react";
import { openuiIntegration } from "@openuidev/assistant-ui";
function OpenUIModelInstructions() {
useAssistantInstructions(openuiIntegration.instructions);
return null;
}
// Render this inside AssistantRuntimeProvider (next step).
export function OpenUIConfigProvider({ children }: { children: ReactNode }) {
const aui = useAui();
const config = AuiConfig({
tools: Tools({ toolkit: openuiIntegration.toolkit }),
});
return (
<AuiProvider extends={aui} config={config}>
<OpenUIModelInstructions />
{children}
</AuiProvider>
);
}
```
</Step>
<Step title="Let the model call the OpenUI tools" source="https://www.openui.com/docs/api-reference/assistant-ui">
present_openui renders complete cards, tables, and charts. prompt_openui renders forms and choices, and completes only when the user submits a result. With an AI SDK runtime, shouldContinueAfterOpenUIPrompt starts the next model step only after a prompt_openui result.
```tsx title="src/components/openui-runtime-provider.tsx"
"use client";
import type { ReactNode } from "react";
import { AssistantRuntimeProvider } from "@assistant-ui/react";
import { useChatRuntime } from "@assistant-ui/react-ai-sdk";
import { shouldContinueAfterOpenUIPrompt } from "@openuidev/assistant-ui/ai-sdk";
// The provider from step 1.
import { OpenUIConfigProvider } from "./openui-config-provider";
export function OpenUIRuntimeProvider({ children }: { children: ReactNode }) {
// The runtime's chat transport forwards both OpenUI tools to your backend.
// Continue once prompt_openui has the user's answer; a finished
// present_openui call is the final response and starts no new step.
const runtime = useChatRuntime({
sendAutomaticallyWhen: shouldContinueAfterOpenUIPrompt,
});
return (
<AssistantRuntimeProvider runtime={runtime}>
<OpenUIConfigProvider>{children}</OpenUIConfigProvider>
</AssistantRuntimeProvider>
);
}
```
</Step>
<Step title="Bring your own components" source="https://www.openui.com/docs/api-reference/assistant-ui">
Call createOpenUIIntegration with your own library so the tool schemas, model instructions, and renderers share one component vocabulary.
```ts title="src/lib/openui-integration.ts"
import { createOpenUIIntegration } from "@openuidev/assistant-ui";
import { createShouldContinueAfterOpenUIPrompt } from "@openuidev/assistant-ui/ai-sdk";
// Your component library, made with createLibrary from @openuidev/react-lang.
import { library } from "@/lib/openui";
// Use openui.toolkit and openui.instructions in place of openuiIntegration.
// The tool names are optional; these replace present_openui and prompt_openui.
export const openui = createOpenUIIntegration({
library,
presentToolName: "show_panel",
promptToolName: "ask_panel",
});
// Pass this to useChatRuntime as sendAutomaticallyWhen when you rename the tools.
export const shouldContinue = createShouldContinueAfterOpenUIPrompt({
promptToolName: openui.toolNames.prompt,
});
```
</Step>
</Steps>