Turn any website into clean data. From your own desktop.

Scrape, crawl, extract, and monitor the web with AI — running entirely on your machine, with your own API keys. One flat price. No credits, ever.

  • Runs on your machine
  • Your keys never leave it
  • Cancel anytime
  • No credits
  • No cloud
  • No meter anxiety

Watch a website become a dataset.

Scrape a page, crawl the rest of the site, extract a schema with your own key, set a watch, then query it from Cursor. All on one desktop, none of it through our servers.

Scraping shouldn’t run on a meter.

Cloud scraping APIs

  • Credits burn on JS-heavy pages
  • Your data transits their servers
  • Developer-only APIs

data path · 4 hops · 2 through their servers

  1. you
  2. their cloudmetered
  3. the site
  4. their cloudmetered
  5. you

Magpie

  • Flat $49 a month, unmetered
  • Your data stays on your machine
  • GUI, CLI and MCP — pick your surface

data path · 2 hops · nothing in between

  1. your machine
  2. the site
  3. your machine

Five engines. One flat price.

Everything runs locally. Extraction and research call the LLM you choose, with your own key.

  • Scrape

    Any page → clean markdown, JSON, or a screenshot

    Magpie’s Scrape screen: a books.toscrape.com product page returned as markdown — 200 OK in 1,253 ms, 1,751 characters.
    No header, no breadcrumbs
  • Crawl

    Whole sites, with live progress

    Magpie’s Crawl screen mid-run on quotes.toscrape.com: 13 of 25 pages, no errors, each fetched URL ticked off as it lands.
  • Extract

    Schema-guided structured JSON with your own LLM keys

    Magpie’s Extract: a six-field schema beside the typed JSON record it returned for the same books.toscrape.com page — GBP, 51.77, in stock, 22, title and UPC — for 552 tokens, $0.0001.
  • Monitor

    Watch pages, get diffs, fire webhooks on change

    A Magpie watch on quotes.toscrape.com/random sending its changes to a local webhook, with the latest diff: one quote and author removed, another added.
  • Research

    Multi-source reports, compiled for you

    Key findings from a Magpie research report on llms.txt, each sentence tagged with the numbered sources behind it — compiled from 86 usable facts for $0.02.

Structured extraction, your schema

Paste a schema or let Magpie infer one; get typed JSON every time.

type: object
properties:
  title:    {type: string}
  price:    {type: number}
  in_stock: {type: boolean}
required: [title, price, in_stock]
Magpie’s Extract: that schema run against a books.toscrape.com page, returning in_stock true, price 51.77 and the title “A Light in the Attic” — 506 tokens, $0.0001.

Scrape behind logins

Capture a browser session once; reuse it for authenticated pages.

log in once  →  session saved locally
             →  scrape behind the login
A sign-in page with email and password fields — the kind of login you go through once in the window Magpie opens.

Built-in MCP server

Your AI agents use the Magpie on your machine. Cursor, Claude, any MCP client. Your keys never leave the box.

  • Cursor
  • Claude
  • any MCP client
Magpie’s Connect an agent tile: the MCP server running on 127.0.0.1, with one-click setup for Claude Code, Claude Desktop and Cursor.

From install to data in three steps.

  1. 01

    Install

    Download the desktop app for Windows, macOS or Linux.

    Magpie on first launch: the engine sidebar — Scrape, Crawl, Search, Watch and Research — beside an empty Overview, every count at zero.
  2. 02

    Point at a URL

    Paste a URL and pick an engine — scrape, crawl, extract, monitor or research.

    Magpie’s Scrape screen with a quotes.toscrape.com author page pasted into the URL bar, Scrape picked over Crawl and Watch, and markdown and json chosen as formats.
  3. 03

    Get data

    Clean markdown, JSON or a screenshot, saved to a local cache you own.

    Prefer a terminal?

    magpie scrape https://example.com

    The same page back in 2,459 ms with 200 OK: Albert Einstein’s author page rendered as clean markdown, with markdown and json tabs beside it.

Magpie vs cloud scraping APIs.

Magpie$49/mo, everythingCloud scraping APIsmetered
Credits neededNoYes
Data pathYour machineTheir cloud
LLM keysYours, localBundled
GUIYesNo
CLI + MCPYesAPI only
PriceFlat $49Metered credits

What people point it at.

  • Lead enrichment

    Turn company websites into clean records for your CRM.

  • Price & competitor monitoring

    Watch pricing pages and get a diff the moment they change.

  • AI/RAG pipelines

    Feed agents and vector stores clean markdown, not raw HTML.

  • Market & SEO research

    Crawl a market’s sites and compile what they say into one report.

  • Content gathering

    Collect articles, docs and listings into one consistent format.

  • Internal tools

    Pipe structured page data into your own scripts over the CLI or MCP.

The privacy story the cloud can’t tell.

Magpie runs where your data already is. There is no Magpie server in the path to trust, breach or subpoena.

  • Data never touches our servers

    Requests go from your machine to the site and back. We are not in the path.

    you → shop.example.com → youmagpie servers: 0 hops

  • SSRF guard

    Requests to internal-network addresses are blocked, so a page can’t turn Magpie on your LAN.

    GET 192.168.1.1/adminblocked · internal address

  • A local cache you own

    Results live on your disk, in a cache you can read, move or delete.

    ~/.local/share/magpie/cache.dbowner: you

  • Every LLM call, visible

    Each call to your model provider shows up in history. Nothing happens off the record.

    extract · 552 tokens$0.0001 · logged

No credits on any plan.

-2 months
  • Core

    $0forever

    Scrape and crawl, free.

    • Scrape to markdown, JSON or screenshot
    • Crawl whole sites
    • No page, credit or request limits
    • Runs entirely on your machine
    • Every update
  • Pro

    $49/mo

    Everything, unmetered.

    • Everything in Core
    • Extract with your own LLM keys
    • Monitor pages, fire webhooks
    • Scrape behind logins
    • Research reports
    • CLI and built-in MCP server
    Get Pro

    Unlocks in-app · Cancel anytime

  • Enterprise

    Custom

    Every seat, managed.

    • Everything in Pro, per seat
    • SSO license activation
    • Silent install via MDM
    • Offline activation for air-gapped machines
    • Priority support and security review
    Talk to us

    Per-seat volume pricing

Compare to credit-metered APIs

Questions, answered.

More in the docs
Is it really unmetered?

Yes. $49 a month covers every engine with no page, credit or request limits. AI features call your own LLM provider, billed by them at their rates.

Do you see my data?

No. Scraping, crawling and extraction run on your machine, and results stay in a local cache you own. Nothing you scrape passes through our servers.

Where do my API keys go?

Nowhere but your machine. Keys are stored locally and sent only to the LLM provider you chose — and every call is visible in your history.

Which OS?

Windows, macOS and Linux.

Do I need to code?

No. Everything works from the desktop app. The CLI and the MCP server are there when you want them.

Refund policy?

It's set out in the refund policy.

Stop renting credits.

One flat price, your machine, your keys.

Download free for Windows, macOS and Linux

Windows · macOS · Linux $49/mo flat, cancel anytime