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

# Introduction

> Shelv converts documents into structured Markdown filesystems. Your agents navigate them with standard Unix tools — no parsing code, no embeddings, no retrieval pipeline.

# What is Shelv?

Shelv converts documents into structured Markdown filesystems your AI agents can actually use. Upload a PDF, get back a browsable file tree — chapters, sections, clauses — that your agent navigates with standard Unix tools like `ls`, `cat`, and `grep`.

## Why agent developers use it

Every agent that works with documents reinvents the same plumbing — parsing PDFs, chunking content, figuring out structure, building retrieval. Shelv eliminates all of it. Your agent gets deterministic file paths instead of probabilistic retrieval. Standard file tools instead of custom parsing code. The document's own hierarchy instead of arbitrary chunks.

## How it works

<Steps>
  <Step title="Upload a document">
    Send any PDF to the Shelv API — contracts, regulations, books, whatever.
  </Step>

  <Step title="We structure it">
    Shelv parses, chunks by semantic structure, and builds a Markdown
    filesystem.
  </Step>

  <Step title="Your agent hydrates it">
    Browse with `ls`, read with `cat`, search with `grep`. Use the MCP server,
    OpenClaw skill, sandbox adapters, archive download, or direct REST reads.
  </Step>
</Steps>

## Example output

```text theme={"theme":{"light":"github-light","dark":"poimandres"}}
my-contract/
├── README.md
├── definitions.md
├── article-01-parties.md
├── article-05-laytime/
│   ├── 5.1-calculation.md
│   └── 5.2-exceptions.md
└── clauses/
    ├── force-majeure.md
    └── arbitration.md
```

## Core capabilities

* Async processing — your agent gets a webhook when it's ready
* Optional templates (`book`, `legal-contract`, `charter-party`, `academic-paper`)
* Optional review mode before finalization
* Signed webhook notifications for lifecycle events
* Multiple access modes — MCP server, OpenClaw skill, sandbox adapters, archive download, REST API, or full JSON tree
* API-first — every operation is a single curl call

## Next steps

<CardGroup cols={2}>
  <Card title="Quickstart" icon="rocket" href="/quickstart">
    Upload a document and retrieve structured files in minutes.
  </Card>

  <Card title="API Reference" icon="code" href="/api-reference">
    Explore endpoint schemas and status-specific behavior.
  </Card>

  <Card title="MCP Server" icon="plug" href="/guides/mcp-server">
    Connect Claude Code, Cursor, or any MCP client to Shelv.
  </Card>
</CardGroup>