Parse once, mount anywhere

The document layer
for AI agents

Upload any document. Get hosted Markdown filesystems your agents mount and navigate with ls, cat, and grep — no embeddings, no retrieval pipeline.

terminal
$ ls /shelv/irs-tax-code/
chapters/ appendices/ definitions/ index.md
$ cat /shelv/irs-tax-code/chapters/03-gross-income.md
# Chapter 3: Gross Income
## § 61. Gross income defined
Except as otherwise provided in this subtitle, gross
income means all income from whatever source derived...
$ grep -r "capital gains" /shelv/irs-tax-code/chapters/
chapters/11-capital-gains.md: ## § 1001. Determination of capital gains...
chapters/03-gross-income.md: ...including capital gains as defined in...

How it works

From raw document to mounted filesystem

Step 01

Ingest

Upload contracts, manuals, regulations, books, and other complex documents.

Step 02

Structure

Shelv maps semantic hierarchy into clean Markdown filesystems with deterministic paths.

Step 03

Mount + operate

Mount hosted shelves into any sandbox at inference time. Your agent reads local files — Shelv manages the rest.

Features

Why agents choose Shelv by default

Deterministic over probabilistic

Agents follow stable file paths and hierarchy instead of top-k retrieval guesswork.

Standard tools, zero glue code

ls, cat, grep. Keep your workflow simple without custom parsing or retrieval plumbing.

Structure preserved

Chapters, sections, clauses, appendices — the document's own shape, not arbitrary chunks.

Async-ready with webhooks

Get lifecycle events as processing completes and plug into production workflows.

Hosted and managed

Shelv stores, versions, and serves your shelves. No infrastructure to run — just mount and go.

Framework-agnostic

OpenClaw skill, MCP server, or raw REST API. Works with any agent stack — zero vendor lock-in.

Developer experience

Simple, powerful API

A few lines of curl is all it takes. Upload a document, wait for the webhook, and your agent can start reading structured files immediately.

  • Give your agent file-tool access to any document
  • Browse files with familiar REST endpoints
  • Retrieve individual files by path — just like a filesystem
  • No embeddings, no vector DB, no retrieval tuning
api-example.sh
$ curl https://api.shelv.dev/v1/shelves/shv_9k2f/files \
  -H "Authorization: Bearer sk_live_..."

[
  { "path": "/contract/summary.md",          "size": 2841 },
  { "path": "/contract/parties.md",          "size": 1205 },
  { "path": "/contract/terms/payment.md",    "size": 3102 },
  { "path": "/contract/terms/termination.md","size": 1847 }
]

$ curl https://api.shelv.dev/v1/shelves/shv_9k2f/files/contract/summary.md \
  -H "Authorization: Bearer sk_live_..."

# Contract Summary
**Parties:** Acme Corp ↔ Widget Inc
**Effective Date:** 2025-01-15
**Term:** 24 months, auto-renewing
...

Integrations

Use Shelv from any agent environment

Mount shelves into sandboxes with adapters, connect via MCP or OpenClaw, or call the REST API directly.

OpenClaw Skill

Ingest documents, monitor processing, and hydrate structured files directly into OpenClaw workflows.

View docs

MCP Server

List shelves, read files, and search content from any MCP-compatible client, including Claude Code, Cursor, and Windsurf.

View docs
Compatible withOpenAIAnthropicClaude CodeCursorLangChainCrewAIWindsurf

Pricing

Simple, transparent pricing

Start free, scale as you grow.

Free Trial

$0

No credit card required

  • 1 document
  • Up to 50 pages
  • 14-day access period
  • Full API access
  • Webhook delivery
Start free
Recommended

Pro

$19/mo

500 pages included

  • Everything in Free
  • 500 pages/month included
  • $0.02/page overage
  • Unlimited documents
  • Priority support
Get started

Estimate your monthly cost

500 pages
$19 base + $0.00 overage = $19.00/mo

Frequently asked questions