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

# Overview

> EverOS OSS is a lightweight, md-first, edge-cloud isomorphic memory framework for AI agents.

EverOS OSS gives AI agents persistent memory without external infrastructure. The core design decision: **Markdown files are the source of truth for memory, not a database.**

A three-component embedded stack handles state and retrieval on top:

```
Markdown (truth)  +  SQLite (state)  +  LanceDB (index)
```

This one decision drives the rest of the architecture — zero external services, human-readable memory files, offline-capable edge deployment, and a cloud migration path that requires no code or data changes.

## How the Design Cascades

```
Markdown-first (md is truth)
├── index rebuildable from md       →  zero external services, no ops overhead
├── md portable + API isomorphic    →  edge prototype migrates to cloud as-is
├── md partitioned by owner dir     →  user / agent / knowledge physically isolated
└── md is a text carrier            →  multimodal content unified under one protocol
```

## Three Memory Tracks

Memory in EverOS OSS is organized into three top-level directories, each answering a different question:

```
memory-root/
├── users/      User Memory    · "Who is this user and what have they said?"
├── agents/     Agent Memory   · "What has this agent learned?"
└── knowledge/  Knowledge Wiki · "What reference material has the system been given?"
```

Writes are routed automatically based on message `role`. No manual routing or configuration required.

## EverOS OSS vs Cloud

|                        | OSS                                  | Cloud                      |
| ---------------------- | ------------------------------------ | -------------------------- |
| Deployment             | Self-hosted                          | Managed                    |
| Storage                | md + SQLite + LanceDB                | Hosted infrastructure      |
| External services      | None                                 | None to configure          |
| Offline / edge capable | Yes                                  | No                         |
| Setup                  | pip install or Docker                | API key                    |
| Data location          | Fully local                          | Yours, on EverMind servers |
| Migration path         | Start here, move to Cloud when ready | —                          |

**Choose OSS** for full data control, edge and offline deployment, or custom infrastructure. **Choose Cloud** for managed scale without ops overhead.

The two share the same API contract. Code written against the OSS embedded library migrates to Cloud by swapping method calls for equivalent HTTP requests — parameters and return structures are identical.

<CardGroup cols={2}>
  <Card title="Quickstart" icon="bolt" href="/open-source/quickstart">
    Install EverOS OSS and run your first memory write.
  </Card>

  <Card title="How Memory Works" icon="cube" href="/open-source/how-memory-works">
    The md-first storage model and embedded stack in detail.
  </Card>

  <Card title="Memory Types" icon="layer-group" href="/open-source/memory-types">
    User, agent, and wiki memory — types, structure, and routing.
  </Card>

  <Card title="Retrieval" icon="magnifying-glass" href="/open-source/retrieval">
    How mRAG selects the right memory at the right granularity.
  </Card>
</CardGroup>