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

# Welcome to StateBase

> Ship AI agents with confidence.

<img className="block dark:hidden" src="https://mintcdn.com/statebase/L5_bhFgOdSrvq9YK/images/hero-brand.png?fit=max&auto=format&n=L5_bhFgOdSrvq9YK&q=85&s=66941f5d71b8c6ba9a9597cdfd8d69bb" alt="StateBase - Ship with confidence" width="1024" height="1024" data-path="images/hero-brand.png" />

<img className="hidden dark:block" src="https://mintcdn.com/statebase/L5_bhFgOdSrvq9YK/images/hero-brand.png?fit=max&auto=format&n=L5_bhFgOdSrvq9YK&q=85&s=66941f5d71b8c6ba9a9597cdfd8d69bb" alt="StateBase - Ship with confidence" width="1024" height="1024" data-path="images/hero-brand.png" />

## Imagine your AI had a memory that actually worked.

Most AI agents today are like goldfish. They are brilliant for a single turn, but they forget who the user is, what happened five minutes ago, and why they were doing a task in the first place.

**StateBase is the reliability layer for these agents.** We give your AI a persistent brain, a detailed history, and the ability to "rewind" time when things go wrong.

<CardGroup cols={3}>
  <Card title="Start Here" icon="rocket" href="/quickstart">
    Go from stateless to production-ready in **2 minutes**.
  </Card>

  <Card title="See Demos" icon="play" href="/demos/hallucination-recovery">
    Watch how we save agents from infinite loops.
  </Card>

  <Card title="Production" icon="shield-check" href="/playbook/when-to-checkpoint">
    How founders scale agents with 99.9% reliability.
  </Card>
</CardGroup>

## Why Reliability Matters

When an agent fails in production—entering a loop or hallucinating—it doesn't just lose a user. It costs you money in tokens and destroys trust in your product.

We replace brittle "prompt engineering" with **State Engineering**.

### What StateBase handles for you:

* **Consistency**: Your agent always knows its "current status."
* **Memory**: It recalls user preferences from months ago, not just the last few messages.
* **Auditability**: You can see every internal thought the agent had before it answered.
* **Safety**: Revert your agent to a "safe" checkpoint if it goes off-rails.

## Built for Engineering Teams

Whether you are using OpenAI, Anthropic, or local models, StateBase acts as the unified persistence layer for your entire agent fleet.

<CardGroup cols={2}>
  <Card title="Minimal SDK" icon="box">
    Pydantic-native, async-ready, and lightweight.
  </Card>

  <Card title="Open Standards" icon="bolt">
    REST-first architecture with full OpenAPI support.
  </Card>
</CardGroup>

## Community & Support

<CardGroup cols={2}>
  <Card title="GitHub" icon="github" href="https://github.com/statebase/statebase">
    Star us on GitHub.
  </Card>

  <Card title="Dashboard" icon="layout" href="https://app.statebase.org">
    Monitor your agents in real-time.
  </Card>
</CardGroup>