> ## Documentation Index
> Fetch the complete documentation index at: https://docs.a21e.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Introduction

> a21e is an AI performance layer. Type what you need, get expert results.

## What is a21e?

a21e turns natural-language instructions into expert AI outputs. Instead of writing prompts, you describe what you want — a21e selects the best techniques, compiles an optimized prompt, routes it to the right model, and returns production-quality results.

```
"Write a database migration to add a soft-delete column to the users table"
```

Behind the scenes, a21e:

1. **Parses your intent** — understands what you're asking for, the constraints, and the success criteria
2. **Selects techniques** — matches your intent to curated prompt strategies from the prompt library
3. **Compiles a prompt** — assembles a structured prompt with your context, preferences, and memory
4. **Routes to the best model** — picks the optimal LLM for the task (or deliberates across multiple models)
5. **Returns results** — delivers the output with quality scoring and feedback capture

## Key capabilities

<CardGroup cols={2}>
  <Card title="Intent execution" icon="bolt" href="/guides/intents">
    Submit natural-language intents via API or the Huddle interface. a21e handles prompt engineering automatically.
  </Card>

  <Card title="Multi-model deliberation" icon="scale-balanced" href="/guides/huddle">
    For complex decisions, a21e runs multiple models in parallel — each plans, critiques the others, and a consensus emerges.
  </Card>

  <Card title="Persistent memory" icon="brain" href="/guides/memory">
    a21e remembers your preferences, corrections, and context across sessions. Every interaction makes future results better.
  </Card>

  <Card title="OpenAI-compatible API" icon="plug" href="/guides/openai-compatibility">
    Drop-in replacement for the OpenAI chat completions endpoint. Switch one line and get enhanced results.
  </Card>
</CardGroup>

## How it works

<Steps>
  <Step title="Get an API key">
    Sign up at [a21e.com](https://a21e.com) and generate an API key from the dashboard.
  </Step>

  <Step title="Submit an intent">
    Send a natural-language instruction to the API. a21e handles the rest.

    ```bash theme={null}
    curl -X POST https://api.a21e.com/v1/rpc \
      -H "Authorization: Bearer YOUR_API_KEY" \
      -H "Content-Type: application/json" \
      -d '{
        "method": "intent.submit",
        "params": {
          "input": "Write unit tests for a React login form component",
          "auto_execute": true
        }
      }'
    ```
  </Step>

  <Step title="Get results">
    a21e returns the executed output along with quality metadata, credits used, and the model selected.
  </Step>
</Steps>

## Pricing

a21e uses a credit-based model. **1 credit = 1 enhancement** (one prompt compilation and execution).

* **Managed mode** — a21e handles LLM calls. Credits cover both prompt engineering and model inference.
* **BYOK mode** — Bring your own API keys for Anthropic, OpenAI, Google, or xAI. Credits only cover prompt engineering.

See [Credits & billing](/account/credits) for details.