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

# Introduction

> The blazing-fast AI gateway. Built in Rust. ~5ms overhead.

# Ship AI Apps. Sleep at Night.

Raptor is a **high-performance AI proxy** that sits between your app and AI providers. Built entirely in **Rust**, it adds enterprise features with just **\~5ms of overhead**.

<CardGroup cols={2}>
  <Card title="~5ms Overhead" icon="bolt">
    Rust-powered performance. Your users won't notice we're there.
  </Card>

  <Card title="60%+ Cost Savings" icon="piggy-bank">
    Semantic caching means you stop paying for the same answer twice.
  </Card>

  <Card title="AI Firewall" icon="shield">
    Block prompt injection and jailbreaks before they hit your model.
  </Card>

  <Card title="Full Audit Trail" icon="clipboard-list">
    Every request logged. Compliance-ready from day one.
  </Card>
</CardGroup>

## The Problem

You're building with AI. You need:

* **Caching** to stop burning money on repeated queries
* **Security** to block prompt injection attacks
* **Logging** for compliance and debugging
* **Speed** because your users won't wait

Building this yourself? That's months of work. And you'll probably get it wrong.

## The Solution

**One line of code. All the features.**

```python theme={null}
client = OpenAI(
    base_url="https://proxy.raptordata.dev/v1",  # Just change this
    default_headers={"X-Raptor-Api-Key": "rpt_..."}
)
```

That's it. Your existing code keeps working. You get caching, security, and logging instantly.

## Why Rust?

We chose Rust because your latency budget matters:

| Component        | Overhead      |
| ---------------- | ------------- |
| Request parsing  | \~0.5ms       |
| Firewall check   | \~2ms         |
| Cache lookup     | \~1ms         |
| Evidence logging | \~0ms (async) |
| **Total**        | **\~5ms**     |

Compare that to Python/Node.js proxies adding 50-100ms. When your AI calls take 500ms, that's the difference between "snappy" and "sluggish."

## Works With Everything

<CardGroup cols={3}>
  <Card title="OpenAI" icon="robot" href="/integrations/openai">
    GPT-4, GPT-4o, o1, embeddings
  </Card>

  <Card title="Anthropic" icon="message" href="/integrations/anthropic">
    Claude 3.5, Claude 3
  </Card>

  <Card title="Any HTTP API" icon="globe">
    If it speaks HTTP, we proxy it
  </Card>
</CardGroup>

## Get Started in 2 Minutes

<Steps>
  <Step title="Sign Up">
    Create a free account at [raptordata.dev](https://raptordata.dev)
  </Step>

  <Step title="Get Your Keys">
    Copy your API key and Workspace ID from the dashboard
  </Step>

  <Step title="Change One Line">
    Point your SDK at `proxy.raptordata.dev` and add your Raptor headers
  </Step>
</Steps>

<Card title="Quickstart Guide" icon="rocket" href="/quickstart">
  Full setup guide with code examples for Python, TypeScript, and cURL
</Card>

## What You Get

| Feature            | Free      | Pro      | Enterprise |
| ------------------ | --------- | -------- | ---------- |
| Requests/month     | 1,000     | 100,000  | Unlimited  |
| Semantic caching   | Yes       | Yes      | Yes        |
| AI firewall        | Yes       | Yes      | Yes        |
| Evidence retention | 7 days    | 30 days  | Custom     |
| Support            | Community | Priority | Dedicated  |

<Note>
  **Free tier is forever free.** No credit card required. Start building now.
</Note>