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

# Introduction

> ModelRoute is a provider-opaque AI execution marketplace. One API, hundreds of models, zero provider lock-in.

## What is ModelRoute?

ModelRoute is an AI model execution platform built for builders. You submit a request with a model slug and input. ModelRoute routes it to the best available provider, executes it, and returns a canonical response. You never see or interact with the underlying provider.

Whether you're a solo developer validating an idea, a startup scaling to millions of users, or an enterprise building mission-critical AI pipelines — ModelRoute gives you a single, reliable API that just works.

## Why ModelRoute?

<CardGroup cols={2}>
  <Card title="One API, Hundreds of Models" icon="shuffle">
    No more managing provider SDKs, API keys, and response formats. One integration handles image generation, video, audio, text — across every provider.
  </Card>

  <Card title="Async-First, Built to Scale" icon="bolt">
    Every execution is asynchronous. Submit your request, get a tracking ID, receive results via webhook. No fragile HTTP connections, no timeouts, no surprises.
  </Card>

  <Card title="12 Deterministic Error Codes" icon="shield-check">
    Every provider error is classified into one of 12 normalized codes with a retryable flag. No more parsing provider-specific error formats. Your error handling code works across every model.
  </Card>

  <Card title="Complete File Isolation" icon="lock">
    Your files are never exposed to providers. Upload to ModelRoute, reference by ID, download results by ID. Provider CDN URLs are never visible to your application.
  </Card>
</CardGroup>

## How it works

<Steps>
  <Step title="Submit your request">
    POST an execution with a model slug and input. ModelRoute validates your request, estimates cost, and places a balance hold — all in milliseconds. You receive a `202 Accepted` with an execution ID.
  </Step>

  <Step title="ModelRoute handles the rest">
    The platform selects the best available provider, translates your input into the provider's format, executes it, and handles retries and failover automatically.
  </Step>

  <Step title="Receive your result">
    Get notified via webhook when your execution completes. Or poll for status while testing. Either way, the result is in ModelRoute's canonical format — consistent across every provider.
  </Step>
</Steps>

## Built for builders

We know that many builders today come from non-technical backgrounds — and that's great. ModelRoute is designed to be a point-and-shoot solution: robust, resilient, and reliable from day one.

No need to understand provider APIs, manage rate limits across vendors, or build retry logic. ModelRoute handles the infrastructure so you can focus on your product.

**For testing and small apps**: use polling to check execution status. It's the simplest path from idea to working prototype.

**For production at scale**: set up webhooks once and forget about it. Results are pushed to your endpoint the moment they're ready — no rate limits, no wasted API calls, no dropped results.

## Key concepts

| Concept             | Description                                                                        |
| ------------------- | ---------------------------------------------------------------------------------- |
| **Execution**       | A single AI model run. Lifecycle: PENDING → PROCESSING → COMPLETED or FAILED.      |
| **Model slug**      | A canonical identifier for a model (e.g., `flux-1.1-pro`). Provider-agnostic.      |
| **File reference**  | A `file_<uuid>` identifier for uploaded files. Used in inputs and outputs.         |
| **Balance hold**    | Cost estimate locked before execution. Settled on completion, released on failure. |
| **Idempotency key** | A unique key to prevent duplicate executions. Same key = same result.              |
| **Webhook**         | Your endpoint that receives execution results automatically.                       |

## Next steps

<CardGroup cols={2}>
  <Card title="Quickstart" icon="rocket" href="/quickstart">
    Run your first execution in 5 minutes.
  </Card>

  <Card title="Authentication" icon="key" href="/authentication">
    API key management, rotation, and permissions.
  </Card>
</CardGroup>