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

# Memory

> Qodex memory stores project facts and learned behavior so the agent does not start from zero every run.

# Memory

Memory is the project knowledge Qodex keeps between sessions. It helps the agent remember your app, auth patterns, workflow rules, API behavior, UI behavior, and previous findings.

Without memory, the agent would rediscover the same context every time it writes or runs tests.

## Memory files

| File                 | Owner         | Loaded every turn | Purpose                                                     |
| -------------------- | ------------- | ----------------- | ----------------------------------------------------------- |
| `project.md`         | You           | Yes               | App overview, stack, accounts, focus areas, and team rules. |
| `workflow_rules.md`  | You and agent | Yes               | Rules the agent must obey on every run.                     |
| `api_behavior.md`    | Agent         | On demand         | API patterns and endpoint behavior.                         |
| `ui_behavior.md`     | Agent         | On demand         | Page behavior and UI observations.                          |
| `findings_digest.md` | Agent         | On demand         | Summary of important findings.                              |

The always-loaded files stay small and universal. The larger behavior files are indexed and read when needed.

## Why memory is markdown

Markdown is editable, inspectable, and easy to curate. Memory is meant to be project documentation for the agent, not an opaque vector database.

This makes it easier to answer: "Why did Qodex know this?"

## What to put in memory

* Stable auth flows and test accounts.
* Routes or pages that need special handling.
* Team rules such as "never run destructive tests on production."
* Known quirks that look like bugs but are expected.
* Product areas that need extra coverage.

Do not put passwords, tokens, API keys, or private customer data in memory.

## Explore this section

<CardGroup cols={2}>
  <Card title="project.md" icon="pencil" href="/memory-project-md">
    Learn what belongs in the main user-managed memory file.
  </Card>

  <Card title="Editing and curating memory" icon="pencil" href="/memory-editing-and-curating">
    Keep memory accurate, compact, and useful.
  </Card>

  <Card title="How memory works" icon="arrow-right-to-bracket" href="/kb-how-memory-works">
    See what is loaded every turn and what is loaded on demand.
  </Card>

  <Card title="Memory concept overview" icon="book-open" href="/concepts-memory">
    Read the shorter conceptual intro.
  </Card>
</CardGroup>

## On the roadmap

<Tip>
  Planned improvements include reflection passes that update agent-managed memory after scans and source-aware memory generated from linked repository routes and components.
</Tip>