proxy.wiki is a reference for how proxies and web scraping actually work. Guides you can follow, definitions you can trust, and measurements we ran ourselves and published with the method.
#Why this site exists
Most writing about proxies is written to sell them. The result is a body of material that explains the categories in marketing terms, repeats figures nobody has verified, and stops exactly where the technical detail begins. If you have ever searched for why a request fails through a proxy and found ten pages of the same paragraph, you have met the problem.
This site is built the other way round. It starts from the mechanism, states what is testable, and shows the test. Where we do not know something, the page says so.
#What you will find here
| Section | What it is for |
|---|---|
| Guides | How to do a specific thing, with commands and code that were run before publication |
| Glossary | Precise definitions of the vocabulary, cross-linked and cited to the relevant specification |
| Research | Measurements we took ourselves, published with the method, the versions and the date |
| Tools | A proxy checker, a connection and leak test, and a cost calculator. Free, no account |
#How the material is written
Three rules govern what may appear on a page, and the publishing system enforces the first one rather than leaving it to memory.
#1. A figure without a source is not published
Every number is either something we measured, in which case the date, the versions and the method appear alongside it, or something a source published, in which case the source is linked. A claim with neither is rejected before the page can go live.
This is why you will not find a table of provider pool sizes here. No independent party has counted them, so any figure we printed would be a vendor’s claim wearing the clothes of a measurement.
#2. A test is described well enough to repeat
Where an article says we measured something, it says how. The listener, the script or the command is printed on the page, together with the software versions and the date. The point is not that you should trust our result; it is that you should be able to get your own.
Software changes. A measurement taken on one version of one client, on one day, is a fact about that version on that day, and our research pages say so explicitly rather than presenting a result as a permanent property.
#3. A gap looks like a gap
Where we hold no verified value you see an em dash. Never a zero, never an estimate, never a plausible-looking guess. The same field is null in our public dataset, and the API response states in plain words that null means unknown rather than none.
#Who it is for
People who have to make something work. Engineers debugging a request that fails only through a proxy, data teams deciding what a category actually buys them, and anyone who has been handed a proxy endpoint and a deadline.
It is not written for beginners in the sense of being simplified. It is written for people who want the mechanism rather than the summary, which is a different thing and often easier to follow.
#What we are not
- Not a proxy provider. We do not sell or resell proxy access.
- Not a certification body. Nothing here is an endorsement, an approval or a guarantee of any product.
- Not legal advice. Several pages touch on what is permitted when collecting data. Those notes describe the landscape; they do not replace advice on your own situation.
- Not finished. The site is young and the coverage is uneven. Sections that do not exist yet are absent rather than filled with placeholder material.
#Machine access
Everything published here is available in machine-readable form, deliberately. There is an index at /llms.txt, a full-text copy at /llms-full.txt, and every article can be requested as Markdown by adding ?format=md to its address. Structured data is emitted as JSON-LD on every page.
We would rather be quoted accurately than crawled awkwardly, so the formats exist and the crawlers are welcome.
#Corrections
If something here is wrong, out of date, or stated with more confidence than the evidence supports, tell us. Send the page and what is wrong with it through the contact page. Substantive corrections are made on the page rather than quietly, and we would rather fix an error than defend it.
The same applies to gaps. If there is a question you expected this site to answer and it does not, that is useful to know, and it is how most of what is here got written.