The Anti-Bot Bypass Economy in Numbers

Home › Guides › The Anti-Bot Bypass Economy

✓ checked 19 October 2026 · sources: npm registry and download API, PyPI and pypistats (September 2026 downloads), GitHub’s public DMCA notice archive (to 16 October 2026) · tools · takedowns · method · download the data

167 npm packages whose own descriptions say they solve CAPTCHAs or get past anti-bot systems were downloaded 1,926,591 times in September 2026. A hand-picked set of 32 similar Python packages on PyPI was downloaded 16,846,647 times. On GitHub, 676 of the 2,371 DMCA takedown notices published in 2026 so far said the code got around a technical protection. In 599 of those (89%), GitHub added a note saying it “did not find sufficient information to determine a valid anti-circumvention claim”.

167 npm bypass packages1.93M npm downloads, Sept16.8M PyPI downloads, Sept37 Tencent WeChat takedowns in 2026

This page describes the tools neutrally and doesn’t link to them. A download count measures installs, including automated builds, not the number of people using a tool or what they use it for. Some of these tools are also used for testing, accessibility and research.

Why this matters now

Websites use anti-bot systems such as Cloudflare, Akamai, DataDome and Kasada, and CAPTCHAs such as Google reCAPTCHA and hCaptcha, to tell people from automated traffic. Scrapers, price trackers, ticket bots and AI agents use bypass tools to get past them. Until recently the dispute was mostly technical. Now website owners are also using copyright law.

Section 1201 of the US Digital Millennium Copyright Act (DMCA) bans getting around “technological measures” that control access to copyrighted works, and offering tools built to do so. In a 14 October column in Communications of the ACM, Berkeley law professor Pamela Samuelson describes how Google and Reddit have sued SerpApi under Section 1201 over Google’s SearchGuard, a system that “aims to prevent unauthorized third parties from conducting automated bulk scraping of Google search results”. She reports that a California court dismissed Google’s complaint but allowed it to be amended, while a New York court found Reddit’s complaint adequately pleaded, and that “more than 20 similar ยง1201 lawsuits are pending against generative AI developers”. SerpApi argued that the DMCA “is not” an anti-scraping law. For the lawsuits, see our Scraping Law Hub.

GitHub, which hosts many of these tools, sees the same claims in its takedown queue. In a 2020 policy post it said that “less than two percent” of the takedowns it processed were based on circumvention claims. It also said it would have every credible Section 1201 claim reviewed by technical experts, and that where a claim is ambiguous it would “err on the side of the developer”.

What the bypass tools do

We sorted the packages into five groups by what their own descriptions say:

  • CAPTCHA solving: Clients for paid solving services (where people or AI models solve CAPTCHAs sent through an API), browser plugins that pass CAPTCHAs to those services, and local solvers that use speech recognition or vision models.
  • Stealth and “undetected” browsers: Modified or patched versions of Chrome, Firefox or automation tools such as Puppeteer, Playwright and Selenium that hide the signals anti-bot scripts look for.
  • Browser-imitating HTTP clients: Libraries that copy a real browser’s TLS and HTTP/2 fingerprint, or answer JavaScript challenges, so scripted requests look like browser traffic.
  • Bypass APIs and SDKs: Clients for commercial services that generate the cookies, tokens or “sensor data” that systems such as Akamai, DataDome or Kasada expect from a real browser.
  • Other tools that mention a bypass: Site-specific scrapers, AI-agent tools (MCP servers) and downloaders that list getting past Cloudflare as one feature.
Groupnpm packagesnpm downloads, Sept 2026PyPI packagesPyPI downloads, Sept 2026
CAPTCHA solving88644,45314915,911
Stealth and “undetected” browsers20595,1421412,570,630
Browser-imitating HTTP clients24618,36933,348,797
Bypass APIs and SDKs1760,812111,309
Other tools that mention a bypass187,81500
Total1671,926,5913216,846,647

npm: every package found by 28 registry searches (14 anti-bot or CAPTCHA names, each with “bypass” and “solver”) whose name, description or keywords claim to bypass or solve one of them, checked by hand. PyPI: a hand-picked list, because PyPI has no search API. PyPI counts exclude mirrors.

  • CAPTCHA solving is the largest npm group by number of packages: 88 of 167. Many are near-identical clients for the same few paid solving services.
  • Downloads are concentrated: the five largest npm packages account for 70% of npm downloads. 101 of the 167 npm packages released a new version in 2026.
  • Python is far bigger: the 32 PyPI packages had 8.7 times the downloads of all 167 npm packages, mostly from stealth browser tools (12.6 million).
  • General browser-imitating libraries aren’t counted unless their own description claims a bypass. curl_cffi, a Python HTTP client that can imitate browsers’ TLS fingerprints, had 33.0 million downloads in September on its own, more than every counted package together.

The largest packages

PackageRegistryGroupDownloads, Sept 2026
wreq-jsnpmBrowser-imitating HTTP clients485,714
cloakbrowsernpmStealth and “undetected” browsers311,201
puppeteer-extra-plugin-recaptchanpmCAPTCHA solving252,537
puppeteer-real-browsernpmStealth and “undetected” browsers169,553
2captchanpmCAPTCHA solving122,818
patchrightPyPIStealth and “undetected” browsers3,816,940
cloudscraperPyPIBrowser-imitating HTTP clients3,324,315
playwright-stealthPyPIStealth and “undetected” browsers2,759,514
undetected-chromedriverPyPIStealth and “undetected” browsers2,039,101
seleniumbasePyPIStealth and “undetected” browsers1,301,745

Which anti-bot systems npm packages name

Named in package name, description or keywordsnpm packagesDownloads, Sept 2026
Cloudflare (any mention)911,479,171
Google reCAPTCHA79952,692
DataDome23584,012
Cloudflare Turnstile47555,562
hCaptcha40513,753
Arkose Labs / FunCaptcha26348,258
GeeTest23242,119
Imperva / Incapsula438,438
Akamai832,424
Kasada430,760
HUMAN / PerimeterX58,484

A package that names several systems is counted under each, so the rows add up to more than the total. Naming a system doesn’t mean the package works against it; we didn’t test any package.

For which websites use these anti-bot systems and CAPTCHAs, see our Anti-Bot Census and CAPTCHA market share pages.

Anti-circumvention takedowns on GitHub

GitHub publishes every DMCA notice it processes in a public archive. Since 2021 its notice form has asked whether the sender claims a technological measure that controls access to their work. Where GitHub couldn’t confirm that part of the claim but acted on other copyright grounds, it adds a note. GitHub says a published notice “does not mean that the content was unlawful or wrong”.

YearNotices publishedClaimed a technological measureGitHub note: no valid anti-circumvention claim found
20211,8247055
20222,321365330
20232,049410396
20242,025471417
20252,463652606
20262,371676637

2026 is to 16 October. Counter-notices and retractions are excluded. Notices sent as letters rather than through GitHub’s form, such as Tencent’s, have no yes/no answer, so they aren’t in the “claimed” column.

The note appeared on 637 of 2,371 notices in 2026 (26.9%), up from 330 of 2,321 in 2022 (14.2%). Circumvention claims have become more common since GitHub’s form began asking about them, and most of the published ones carry the note.

Tencent and WeChat chat-history tools

Lawyers for Tencent and WeChat International filed 37 notices in 2026, the first on 8 January and the latest on 13 October. We found no notices from WeChat International in the archive before 2026. The notices target tools that find WeChat’s database encryption keys to decrypt or export users’ chat histories (33 of the 37 use the words “decrypt” or “chat history”). All 37 cite Section 1201, and all quote WeChat’s terms banning access “through the use of any automated bots, software, engines, crawlers, scrapers, data mining tools or the like”.

Together they name 65 repositories. In 10 cases GitHub says it processed the notice against the repository’s entire network of copies (forks) because the network had more than 100 repositories, which covers about 13,600 repositories in total (GitHub gives one network size only as “3.2k”). These are tools for reading WeChat data, not general anti-bot bypass tools. They are included because the notices use the same anti-bot language and the same law.

Kasada

Kasada, an anti-bot company, filed 6 notices between March 2025 and May 2026 (2025-03-17, 2025-05-30, 2026-01-20, 2026-03-23, 2026-04-01, 2026-05-07). They target repositories with names such as Kasada-Reverse and Kasada-Solver, and one npm package, which Kasada says reproduce its protected JavaScript. 5 of the 6 said Kasada used technological measures. On the last 3, GitHub added its note that it couldn’t find a valid anti-circumvention claim but acted on other copyright claims. Earlier, Akamai filed 7 notices between 2018 and 2021 and Imperva filed 1 in 2023, going by notice file names.

What these numbers can’t show

  • Downloads aren’t users. Automated builds, test pipelines and mirrors add downloads. One project can install the same package thousands of times.
  • A claim isn’t a result. We counted what packages say about themselves. We didn’t install or test any of them.
  • The PyPI list is hand-picked (46 names checked), so it covers the best-known packages rather than all of them.
  • Takedown notices are allegations. GitHub publishes the notices it processes, so notices it didn’t act on aren’t counted here.

Method

On 19 October 2026 we ran 28 npm registry searches (up to 500 results each) and kept packages whose name, description or keywords named an anti-bot system or CAPTCHA together with a word such as bypass, solve, solver, evade or undetected. That gave 174 candidates. We checked each one by hand and removed 7 that were false matches or tools for site owners, such as a token verifier. Download counts are for 1 to 30 September 2026 from the npm downloads API. On PyPI we checked 46 candidate names and kept those whose own name, summary, keywords or description title claims a bypass, CAPTCHA solving or undetected/stealth operation. Downloads (excluding mirrors) are summed from pypistats for September 2026. GitHub figures come from a full copy of the github/dmca repository at commit 26dbfcf (16 October 2026), with counter-notices and retractions removed by file name.

Download the data

  • npm packages (CSV): every checked package, its group, systems named, September downloads and latest release date, plus the excluded packages and why.
  • PyPI packages (CSV): included and excluded packages with September downloads.
  • GitHub DMCA counts (CSV): yearly totals, plus every Tencent WeChat and Kasada notice with a link to the notice.

Cite as: ProxyPicker, “The Anti-Bot Bypass Economy in Numbers”, 19 October 2026, https://proxypicker.com/anti-bot-bypass-economy/.

Related: Anti-Bot Census · CAPTCHA market share · Scraping Law Hub.

Press & data requests: email [email protected] for interviews, the raw data behind this page, or corrections. See also our contact page.