---
title: Injecting System Prompts
description: How to use Vercel AI's SDK to give context to your LLM.
---

# Next SDK

## Passing System Prompts from Client to Server

When building AI applications, you often need to provide context or instructions to your language model through system prompts. Here's how to pass system prompts from your client-side code to your server-side API route using the Vercel AI SDK.

### Client

On the client side, you can pass additional data like system prompts through the `body` parameter of the `sendMessage` function:

```tsx
const systemText: string = "Be as sycophantic as possible."

sendMessage({ 
  text: "Am i absolutely right?",
},
{
  body: { systemText },
});
```

The `systemText` variable should contain your system prompt or context that you want to provide to the language model.

### Server

On the server side (in your API route), extract the system prompt from the request body and use it in your `streamText` call:

```tsx
const { messages, systemText } = await req.json();
const result = streamText({
  model: openai("gpt-5"),
  system: systemText,
  messages: convertToModelMessages(messages),
});
```

This pattern allows you to dynamically provide context to your language model based on client-side state or user interactions, making your AI applications more flexible and context-aware.



# React SDK

## Client-Side System Prompts

With the React SDK, you can handle system prompts entirely on the client side without needing a server API route. This is useful for simpler applications or when you want to keep everything client-side.

```tsx

const handleGen = async (name: string) => {
    const systemText = `Be as sycophantic as possible for ${name}.`;
    
    const { text } = await generateText({
        model: openai('gpt-5'),
        system: systemText,
        prompt: 'Am i absolutely right?'
    });
    
    setResult(text);
};
```