> ## Documentation Index
> Fetch the complete documentation index at: https://docs.e6data.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Memory and context

> The kinds of state AgentFlow keeps around the model, and where each lives.

export const DocMeta = ({type, applies, owner}) => {
  const LABELS = {
    tutorial: "Tutorial",
    "how-to": "How-to guide",
    explanation: "Explanation",
    reference: "Reference"
  };
  return <div className="e6-meta">
      <span className={type === "how-to" || type === "tutorial" ? "e6-eyebrow" : "e6-eyebrow e6-eyebrow--quiet"}>{LABELS[type] || type}</span>
      {applies ? <span>Applies to: {applies}</span> : null}
      {owner ? <span>Owner: {owner}</span> : null}
    </div>;
};

<DocMeta type="explanation" applies="All deployments" />

AgentFlow's runtime, the harness around the model, handles four classes of state:

| State | What it includes |
| - | - |
| Conversation and run state | Messages, task progress, tool calls, step outputs, checkpoints and artifacts |
| Project context | Your description, business rules, glossary and files |
| Semantic-layer context | Selected data, descriptions, statistics, business context and indexes |
| Reusable execution knowledge | Skills, workflow definitions and memoized recipes |

## Where it lives

Redis coordinates live activity. Configured durable storage keeps product assets and history. See [Architecture](/agentflow/concepts/architecture).

Because of this split, a client can disconnect and reattach to an active chat stream, and a direct workflow run continues as a detached run. A paused run, such as one [waiting for human input](/agentflow/automate/human-input), resumes from durable state.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.