Marreta

Open-source URL cleaner and paywall bypass tool, honestly reviewed. What you get when you run your own instance instead of trusting someone else’s.

TL;DR

• What it is: GPL-3.0 open-source tool that strips paywalls, removes tracking parameters, cleans HTML, and returns a readable version of blocked articles [README].
• Who it’s for: Anyone who hits paywalls regularly and wants a self-hosted instance they control — rather than relying on public proxies that disappear without warning.
• Cost: Free to self-host on any Docker-capable VPS. A public instance at marreta.pcdomanual.com handles open access at no charge.
• Key strength: Unusually rich integration layer for a ~1,100-star project — Firefox extension, Chrome extension, Telegram bot, Bluesky bot, and Apple Shortcuts, all built by the community [README].
• Key weakness: Third-party English-language coverage is thin. The project originated in Brazil and documentation skews Portuguese-first. And the entire category — paywall proxies — sits in a legal gray zone you should think about before running a public instance.

---

What is Marreta

“Marreta” is Portuguese for sledgehammer. The tagline — “Chapéu de paywall é marreta!” — roughly translates to “A sledgehammer for paywalls.” The public instance has processed over 88,000 blocked pages as of this writing, according to the homepage counter.

The mechanics are straightforward: you paste a paywalled or tracking-laden URL into Marreta, and it fetches a cleaned version by switching user agents (pretending to be a search engine crawler, which publishers typically let through to preserve SEO indexing), stripping tracking parameters, fixing relative URLs, removing unwanted page elements, and caching the result. You get a readable article back. The publisher gets the same impression Google would get.

It was created by the team behind Manual do Usuário, a Brazilian tech publication, and sits at 1,139 GitHub stars with GPL-3.0 licensing. It’s PHP 8.4+ under the hood, deployed via Docker Compose, and has optional Selenium integration for JavaScript-heavy pages that won’t render without a real browser engine.

Unlike 12ft.io — the SaaS tool this category of software essentially replaced — Marreta is something you run yourself. You own the instance. You control what domains are blocked, what caching backend you use, and who has access.

---

Why people choose it

Coverage in English-language media is nearly absent. A January 2025 self-hosted newsletter mentioned the project in passing [1]. The AlternativeTo pages for similar tools — 13 Feet Ladder and Ladder — confirm there’s a small but active community around self-hosted paywall proxies, but Marreta doesn’t appear on those pages directly [2][3].

What you find instead is real usage: the public instance counter, a Telegram bot with its own following, and community-built browser extensions distributed through the official Firefox Add-ons and Chrome Web Store channels. For a Brazilian project with primarily Portuguese documentation, reaching 1,100+ GitHub stars is not nothing.

The reasons people choose Marreta over the alternatives come down to three things:

Reliability ownership. Public instances of 12ft.io, smry, and similar tools have come and gone. When you self-host Marreta, you’re not dependent on someone else’s uptime or decisions. The same logic that applies to every self-hosted tool applies doubly here, because paywall bypass tools attract DMCA pressure that can shut down public services fast.

Browser integration that actually works. The Firefox and Chrome extensions (built by Clarissa Mendes, an external contributor) let you click a toolbar button on any paywalled page and get redirected to your own Marreta instance. That removes the copy-paste friction that makes most self-hosted proxy tools annoying in daily use [README].

Multi-platform reach. Most comparable tools are web-only. Marreta has a Telegram bot, a Bluesky bot, and an Apple Shortcuts integration — all community-built, all pointing at either the public instance or your own [README]. That’s four distribution channels for a ~1,100-star project.

---

Features

Based on the README and public instance:

URL processing:

• Automatic URL cleaning and normalization [README]
• Tracking parameter removal (UTM and similar) [README]
• Forced HTTPS [README]
• Relative URL correction [README]
• User agent rotation to avoid publisher blocks [README]

Content handling:

• HTML cleaning and optimization [README]
• Custom CSS and JavaScript injection (configurable per-domain rules) [README]
• Unwanted element removal (popups, banners, cookie notices) [README]

Infrastructure:

• Caching layer with S3 backend support [README][wiki]
• Selenium Hub integration for JavaScript-heavy pages (Chrome and Firefox) [README][wiki]
• PHP-FPM + OPcache for performance [README]
• HTTP proxy support [README]
• Domain blocklist and DMCA compliance tooling with custom messages [README]
• Configurable per-request headers and cookies [README]
• REST API [feature list]
• Plugin architecture [feature list]

Integrations (all community-built):

• Firefox extension (Chrome Web Store + addons.mozilla.org) [README]
• Chrome extension [README]
• Telegram bot [README]
• Bluesky bot [README]
• Apple Shortcuts [README]

Language support: Portuguese (Brazil), English, Spanish, German, Russian [README]

What’s missing compared to more mature tools: there’s no web UI for managing domain rules beyond editing config files, no user authentication layer on the proxy itself (relevant if you’re running a semi-public instance), and the Selenium integration — required for JS-heavy paywalls — needs to be set up and maintained separately.

---

Pricing: self-hosted vs. public instance math

Marreta doesn’t have SaaS pricing because it isn’t a SaaS. This is the tool you run instead of paying for or depending on a hosted service.

The relevant comparison is against the public instance at marreta.pcdomanual.com: it’s free to use, but you’re trusting someone else’s availability, someone else’s DMCA decisions, and someone else’s uptime. If the maintainers pull it down — or get a takedown request from a major publisher — your workflow breaks.

Self-hosting costs:

• Software: $0 (GPL-3.0) [README]
• Minimal VPS (2 vCPU, 2GB RAM): $4–6/mo on Hetzner or Contabo
• Domain name (optional, for clean URLs): $10–15/yr
• S3 storage for cache (optional): $1–3/mo at low volume

Without Selenium, a $5 VPS is sufficient. With Selenium running Chrome or Firefox alongside the PHP app, budget for 4GB RAM minimum — Selenium is memory-hungry.

What you’re saving: not a specific dollar amount from a paywalled SaaS, but the operational risk of relying on someone else’s public instance. If you use this tool daily and wire it into a browser extension pointed at a third-party URL, you’re building a workflow dependency on infrastructure you don’t control. Self-hosting removes that.

---

Deployment reality check

The install path is Docker Compose. Download the compose file, edit four environment variables, run docker compose up -d. That’s the advertised experience [README], and for a basic instance, it’s accurate.

Environment variables you’ll configure:

• SITE_NAME — your instance name
• SITE_DESCRIPTION — shown on the UI
• SITE_URL — full URL including port if non-standard
• SELENIUM_HOST — optional, needed for JS-heavy sites
• LANGUAGE — one of pt-br, en, es, de-de, ru-ru

What can complicate the setup:

The browser extensions need to be configured to point at your instance URL. This isn’t hard, but it’s a manual step that the documentation doesn’t walk you through in detail for the English audience.

Selenium is a separate container setup entirely. If you need to bypass paywalls on JavaScript-rendered pages (which is increasingly common — major publishers like the New York Times render their paywalls client-side), you need Selenium Hub with Chrome or Firefox workers alongside Marreta. That’s a separate wiki page and meaningfully more memory overhead [README][wiki].

S3 caching requires configuring a bucket and credentials. The default in-container cache works for personal use but won’t survive container restarts [wiki].

There’s also the conflict the homepage itself flags: Marreta and ad blockers can interfere with each other, producing blank pages. The recommended workaround is running Marreta in private/incognito mode or temporarily disabling your ad blocker. For daily use, this is friction worth knowing about.

Realistic time estimates: 30 minutes for a basic working instance with an existing VPS and domain. 2–3 hours if you’re setting up a VPS from scratch, configuring reverse proxy, and getting the browser extensions talking to your instance. Add another hour if you want Selenium.

---

Pros and cons

Pros

• GPL-3.0 licensed. You can read the source, fork it, modify it — and the license ensures derivatives stay open [README].
• Unusually broad integrations for its size. Browser extensions, Telegram, Bluesky, Apple Shortcuts. Most tools in this space are web-only [README].
• Multilingual UI. Portuguese, English, Spanish, German, Russian support out of the box — rare for a tool that started as a Brazilian project [README].
• DMCA tooling built in. Domain blocklist and custom DMCA messages let you run a semi-public instance with some legal protection built into the config [README].
• Active public instance. 88,000+ pages processed proves this isn’t vaporware — someone is running it and using it at real scale.
• S3 cache support. Means you can run a shared instance without cache filling up local disk [wiki].

Cons

• Portuguese-first documentation. The README is available in English, but the wiki, community, and most discussion happens in Portuguese. If you hit a setup problem and search for help, you’ll mostly find Portuguese-language resources [README].
• Limited English-language coverage. One newsletter mention [1], no dedicated reviews found in English. You’re making a bet with thin third-party validation.
• Smaller community than alternatives. Ladder has 6,299 stars [3], 13 Feet Ladder has 4,147 [2], Marreta has 1,139. Not a death sentence, but it means fewer people hitting and fixing edge cases.
• Legal gray zone. Paywall bypass tools operate in a category that publishers actively litigate. Running a public instance is riskier than running a private one. The DMCA tooling helps, but it doesn’t make the underlying activity unambiguously legal in all jurisdictions.
• No auth layer. Marreta doesn’t include user authentication out of the box. Anyone who knows your instance URL can use it. If you want access controls, you add them at the reverse proxy layer yourself.
• Selenium is not bundled. JS-heavy paywalls require a separately configured Selenium Hub. It works, but it’s not a one-compose-file deployment anymore.
• PHP runtime. Not a dealbreaker, but Go-based alternatives like Ladder will use less memory for the same workload [3].

---

Who should use this / who shouldn’t

Use Marreta if:

• You hit paywalls regularly enough that it’s worth running your own instance.
• You want browser extensions that point at infrastructure you own.
• You’re in a multilingual environment and the Spanish/German/Russian UI support matters.
• You want the DMCA blocklist tooling to run a semi-public instance for a small team.

Consider Ladder instead if:

• You want more stars, more forks, and a Go runtime that uses less memory [3].
• English-language documentation and community is important to you.
• You’re already comfortable with Go-based tooling.

Consider 13 Feet Ladder instead if:

• You want MIT licensing instead of GPL-3.0 [2].
• Simplicity over features is your priority.

Don’t self-host any of these if:

• You’re not comfortable with the legal implications of running a paywall bypass proxy, even privately.
• You want something maintained by a company with legal counsel — none of these tools have that.
• You hit paywalls occasionally enough that the public instance at marreta.pcdomanual.com is fine.

---

Alternatives worth considering

Ladder (kubero-dev) — GPL-3.0, Go, 6,299 stars. Described as a self-hosted 1ft.io/12ft.io replacement. Inspired by 13ft. More stars than Marreta, more active in English-language communities [3].

13 Feet Ladder (wasi-master) — MIT, Python, 4,147 stars. Pretends to be GoogleBot to access content. Simpler than Marreta, MIT license is more permissive than GPL-3.0 [2].

12ft.io — The SaaS original that all these tools replaced. Spotty availability, no self-hosted option, not recommended if reliability matters.

Unpaywall — Browser extension specifically for academic papers and research. Not useful for news paywalls but worth knowing if that’s your use case.

smry — Summarizer-first approach. Doesn’t try to show the full article, just a summary.

For a non-technical user, the honest answer is: use the public Marreta instance at marreta.pcdomanual.com or 12ft.io until they go down, then decide if it’s worth setting up your own. For someone who already runs a VPS and is comfortable with Docker Compose, Marreta is a 30-minute install with browser integration that works.

---

Bottom line

Marreta is a focused, well-integrated tool doing a specific thing: cleaning URLs and bypassing paywalls via proxy. It’s not trying to be a general-purpose web scraper or a research platform. The browser extensions and multi-platform bots are genuinely impressive for a project this size, and the DMCA tooling shows the maintainers have thought about the legal surface area of running a public instance. The weaknesses are real — thin English coverage, Portuguese-first community, smaller GitHub footprint than alternatives, no bundled auth — but none of them are fatal for someone self-hosting a private instance. If you’re already paying for a VPS and you’d use this daily, it’s worth the 30-minute setup. If you’re looking for the most-maintained tool in this category with the largest English-speaking community, Ladder has more stars and more activity.

---

Sources

1. selfh.st / Ethan Sholly — “This Week in Self-Hosted (10 January 2025)”. https://selfh.st/weekly/2025-01-10/
2. AlternativeTo — “13 Feet Ladder: A site similar to 12ft.io, but is self-hosted”. https://alternativeto.net/software/13-feet-ladder/about/
3. AlternativeTo — “Ladder: Alternative to 12ft.io. Bypass paywalls with a proxy ladder”. https://alternativeto.net/software/ladder-1/about/

Primary sources:

• GitHub repository and README (English): https://github.com/manualdousuario/marreta (1,139 stars, GPL-3.0)
• Official website and public instance: https://marreta.link / https://marreta.pcdomanual.com
• Marreta browser extension (Firefox): https://addons.mozilla.org/pt-BR/firefox/addon/marreta/
• Marreta browser extension (Chrome): https://chromewebstore.google.com/detail/marreta/ipelapagohjgjcgpncpbmaaacemafppe
• Selenium Hub wiki: https://github.com/manualdousuario/marreta/wiki
• Docker Hub: https://hub.docker.com/r/manualdousuario/marreta