> ## Documentation Index
> Fetch the complete documentation index at: https://docs.framework.vibeworkapp.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Prompt Caching

> Reduce API costs by up to 90% with automatic prompt caching

## Overview

Prompt caching reuses previously processed prompt content, reducing API costs and improving latency. The SDK **enables this automatically** for all agents.

<Info>
  Prompt caching is enabled by default. No configuration needed.
</Info>

## Cost Savings

* **Cache read tokens**: 10% of regular price (90% discount)
* **Cache write tokens**: 25% more than regular price
* **Net effect**: Savings increase with conversation length, reaching 75%+ for 10+ turn conversations

## How It Works

The SDK automatically adds cache breakpoints at three locations:

1. **System Prompt** -- your agent's instructions (static)
2. **Tool Definitions** -- all tool schemas (rarely change)
3. **Conversation History** -- the growing message history (last content block)

Cache entries are valid for **5 minutes**. Each cache read refreshes the timer.

### When Caching Helps Most

* **Long conversations**: More turns = more savings
* **Many tools**: Large tool definitions cached once
* **Batch processing**: Multiple similar requests in same session
* **Active sessions**: Continuous conversation within 5 minutes

## Usage

Caching is enabled by default:

```rust theme={null}
let config = AgentConfig::new("You are helpful")
    .with_tools(tools);  // Caching automatically enabled

let agent = StandardAgent::new(config, llm);
```

To disable (not recommended):

```rust theme={null}
let config = AgentConfig::new("You are helpful")
    .with_prompt_caching(false);
```

## Monitoring

Enable debug mode to see cache metrics:

```rust theme={null}
let config = AgentConfig::new("You are helpful")
    .with_debug(true);
```

Output:

```
Cache creation tokens: 5000
Cache read tokens: 12000
Input tokens: 500
```

Access metrics programmatically from the LLM response:

```rust theme={null}
let response = llm.send_message(...).await?;

println!("Cache creation: {}", response.usage.cache_creation_input_tokens);
println!("Cache read: {}", response.usage.cache_read_input_tokens);
println!("Regular input: {}", response.usage.input_tokens);
```

## Limitations

* **Minimum cacheable prefix**: 1024 tokens
* **Cache expiry**: 5 minutes of inactivity
* **Dynamic prompts**: Changing the system prompt or tools between calls invalidates the cache
* **Short conversations**: Overhead may exceed savings for single-turn interactions

<Tip>
  Keep system prompts static and tool sets consistent to maximize cache hits. Avoid including timestamps or dynamic content in your system prompt.
</Tip>

## Next Steps

<CardGroup cols={2}>
  <Card title="Streaming & History" href="/features/streaming">
    Real-time output patterns
  </Card>

  <Card title="AgentConfig API" href="/api/agent-config">
    All configuration options
  </Card>
</CardGroup>