Bespoke work
The scrapes that don't fit in a browser.
A portal you are a customer of. A source that changes shape every month. Thirty competitors, priced every morning before you get in. We build those as projects — and tell you before you pay if it can't be done properly.
HOW IT WORKS You describe the awkward part. We build the thing that runs without you, and put the result where you already look.
What we take on
The extension is deliberately limited to what your browser has already rendered. These are the jobs on the other side of that line.
Systems you already have access to
Your own dashboards, a supplier portal you are a customer of, an internal tool with no export button. Work we do with your authorisation and your credentials, never around anybody's access controls.
Awkward, fragile sources
Aggressive rate limits, markup that changes shape every few weeks, data that arrives in an inconvenient shape. These need something sturdier than a content script, and monitoring so you hear about a break from us rather than from a blank spreadsheet.
Scrapes on a schedule
Every morning, every Monday, every hour. Prices, stock, listings, jobs, tenders, rankings — anything whose value is that it is current. You stop remembering to run it.
Volume, and many sites at once
Hundreds of thousands of rows. Or the same fields taken from thirty competitors, deduplicated and reconciled into one table with a consistent schema, which is usually the harder half.
Delivered where you actually work — Sheets, or anywhere else
A Google Sheet that refreshes itself is the one people ask for most: you open the tab you already had open, and this morning's rows are in it. Also Excel on SharePoint or Drive, rows appended to Postgres or BigQuery, a JSON endpoint you can poll, a webhook into Slack or Zapier, or a file dropped in S3. If it takes an API key or an OAuth consent, that is our side of the job, not yours.
The extension deliberately cannot do this. Writing into your Google account needs an OAuth token and somewhere to keep it, which means a server and an account — the two things it promises never to have. So it lives here instead, where there is a server and you know about it.
One-off rescues
A single list you need this week, cleanly, without learning a tool. Often the cheapest thing on this page.
What to send
Four things. It is usually enough for a yes, a no, or a number.
- The site or sites — a URL. If it needs a login, say so; you don't have to send credentials to get a quote.
- What you need out of it — which fields, and what the finished file or feed should look like. A screenshot of the page with things circled works perfectly.
- How often — once, daily, hourly, or "whenever it changes".
- Your budget and deadline — a figure you have in mind is enough. It tells us immediately whether the scope fits, which saves us both a fortnight of back-and-forth.
Email support@magicscraper.app. Nothing you send is shared with anyone, and quoting costs nothing.
What we'll tell you for free
Including the answers that lose us the job.
If a site can't be collected lawfully or reliably, we say so and explain why. That is worth more to you than the invoice is to us, and it is the reason to ask us rather than whoever said yes to everything.
Whether data may be collected from a given site is a question of that site's terms, its
robots.txt, copyright and database rights, and — where the rows are
about people — data-protection law. We will not build something that bypasses
access controls or that hammers infrastructure to the point of damage.
We'll also tell you when the free extension already does it. Plenty of jobs that feel like projects are ten minutes with the Chrome extension and no invoice at all.
Reasonable questions
What does a project cost?
It depends almost entirely on how hard the site fights and how often it has to run. A single list from a cooperative site is small. A logged-in portal with bot protection, collected hourly and reconciled against two other sources, is not. Tell us the budget you have in mind and we'll tell you honestly whether the scope fits it.
How long does it take?
A one-off extraction is usually days rather than weeks. A scheduled pipeline takes longer because it has to survive the site changing — that is most of the work, and skipping it is why scrapers rot.
Who owns the data and the code?
The data is yours the moment it is collected. Code ownership is agreed per project; say what you need and it goes in writing before anything starts.
What happens when the site changes?
It will. Scheduled work includes monitoring so a break is noticed and fixed, rather than discovered weeks later by someone reading a stale spreadsheet. Terms are agreed per project.
Will you scrape personal data?
Only where there is a lawful basis for it, and we'll ask about that early. If the rows are about identifiable people, that question decides whether the project happens at all.