After (Fortress)
Fortress fixes the browser fingerprint inside Chromium's C++ so your Playwright and Puppeteer code reads as an ordinary Chrome install, while CloakBrowser is an anti-detect browser whose published specs, pricing, and benchmark results we could not verify on this page.
Introduction and methodology note
Both tools exist for the same job: getting scrapers and browser agents past bot protection without a pile of stealth plugins. Fortress is Tilion's open-source Chromium fork. It recompiles Chromium's C++ to correct fingerprints natively, so there are no JavaScript patches to inject and no plugins to maintain. CloakBrowser is an anti-detect browser aimed at the same problem.
They diverge on how you get them, how they are licensed, and what evidence backs the stealth claims. Fortress ships as a pip package, an npm package, a Docker image, a single binary, and a .deb, and its 92-site benchmark is reproducible from the repo. CloakBrowser's distribution model, pricing, and performance claims are not published on any page we reviewed for this comparison, so those cells are marked "not published" rather than guessed.
Methodology: every Fortress figure on this page comes from Tilion's own published pages, linked inline, or from the Fortress repository's benchmark methodology. CloakBrowser figures are stated as plain text from what is publicly known; where nothing is published, we say so. All stealth and accuracy claims below are self-reported by the respective vendors. Test both engines on your own target sites before committing; protected-site success rates vary enormously by site list, IP reputation, and workload.
TL;DR
- Fortress publishes its numbers: 86.4% of 92 protected sites loaded headless on datacenter IPs with no proxies (28 September 2026, reproducible harness in the repo), versus Camoufox at 74.5% and stock Chrome at 34.2%.
- Fortress v3 is free for developers with no caps on machines, sessions, or concurrency; production keys for larger organizations are priced at fortress.tilion.com/pricing. CloakBrowser's pricing is not published on any page we reviewed.
- If you need a verified, reproducible stealth engine that plugs into existing Playwright and Puppeteer code over CDP, Fortress is the safer pick. If you need a feature we cannot verify CloakBrowser offers, we cannot tell you, because its specs are not published.
Quick-comparison table
| Dimension | Tilion Fortress | CloakBrowser | |---|---|---| | Category | Open-source Chromium fork (engine) | Anti-detect browser (not verified in detail) | | Fingerprint approach | 81+ patches in Chromium's C++ layer | Not published | | Install | pip, npm, Docker, .deb, single binary | Not published | | CDP endpoint | Raw CDP on :9222 via Docker | Not published | | Playwright/Puppeteer support | connect_over_cdp, no plugins needed | Not published | | Headless vs headed fingerprint | Same fingerprint in both modes, per vendor | Not published | | Protected-site success rate | 86.4% of 92 sites (self-run, 28 Sep 2026) | Not published | | Free tier | v3 free for developers, no caps | Not published | | Production pricing | Published at fortress.tilion.com/pricing | Not published | | License | Fortress Source Available License 1.1 (v3); older code BSD-3 | Not published | | MCP server for agents | 29 local tools, free, no account | Not published | | Benchmark methodology | Public, reproducible (docs/BENCHMARK.md) | Not published |
Granular differentiator table
The real differentiator in this pairing is verifiability. Fortress publishes its engine internals, its license terms, and a benchmark you can rerun yourself. CloakBrowser publishes none of that on the pages we reviewed.
| Differentiator | Tilion Fortress | CloakBrowser | |---|---|---| | Fingerprint patch count | 81+ patches in Chromium C++ (vendor claim) | Not published | | Fingerprint coherence | Each launch mints a distinct internally coherent machine: GPU, screen, cores, timezone, language, fonts agree | Not published | | Canvas/Audio distinctness | Distinct 40/40 across 40 launches (self-run) | Not published | | Check passes | CreepJS, BrowserScan, rebrowser, WebGL; passes Akamai _abck where stock Chrome gets Access Denied | Not published | | Human-like input | Mouse movement modeled on recordings of about 10,000 people | Not published | | Benchmark harness | 92 sites, two reps, fresh cloud machine per run, 1,584 machines total, methodology in repo | Not published | | Update cadence | Monthly, current major version v3 on Chromium 153.0.8010.36 | Not published | | Agent integrations | browser-use, crawl4ai, Stagehand, LangChain, MCP with 29 tools | Not published |
To be plain about it: on every row above, CloakBrowser either does not publish the figure or we could not verify it. That is not proof CloakBrowser lacks these capabilities; it is proof you cannot evaluate it the way you can evaluate Fortress. If CloakBrowser's site publishes specs or a benchmark, rerun this comparison against those numbers.
Why teams choose Tilion Fortress over CloakBrowser
#### The stealth fix lives in C++, not in JavaScript
Fortress recompiles Chromium's C++ so the fingerprint is correct at the source, per the Fortress product page. JavaScript-level patches can be probed by anti-bot scripts that look for the patch itself; a C++-level fix leaves no injected JS to detect. Fortress passes CreepJS, BrowserScan, rebrowser, and WebGL checks, and passes Akamai's _abck cookie check where stock Chrome gets Access Denied. CloakBrowser's fingerprint mechanism is not published, so you cannot compare approaches, only outcomes.
#### The benchmark is reproducible, not a marketing number
The 92-site test (28 September 2026) ran Fortress headless on datacenter IPs with no proxies against Cloudflare, DataDome, Akamai, HUMAN, Kasada, Imperva, AWS WAF, Arkose, and reCAPTCHA-protected sites: two reps per site, one fresh cloud machine per run, 1,584 machines total. Fortress served 86.4% of pages (95% Wilson interval 81 to 91), Camoufox 74.5%, puppeteer-extra stealth 72.3%, stock Chrome 34.2%. The methodology and harness live in the Fortress repository under docs/BENCHMARK.md, so you can rerun it. No equivalent reproducible benchmark is published for CloakBrowser.
#### It drops into code you already have
Existing Playwright and Puppeteer code connects over CDP with one line: chromium.connect_over_cdp(...). No stealth plugins, no fingerprint patch scripts, no fork of your automation stack. The quick start is pip install tilion-fortress or docker run --rm -p 9222:9222 tilion/fortress:latest. If your team has already written automation against vanilla Chromium, the migration cost from any anti-detect browser, CloakBrowser included, is close to zero.
If you are evaluating engines today, download Fortress and point your existing scraper at it before you commit to anything; the Docker image gets you a testable CDP endpoint in one command.
Best use cases for Tilion Fortress
#### Scraping protected sites with existing Playwright code
If your scraper already runs on Playwright or Puppeteer and keeps hitting 403s or Cloudflare interstitials, Fortress is the direct fix: connect over CDP and the fingerprint problem is handled in the engine. The 86.4% result above was achieved with datacenter IPs and no proxies, which matters if your budget rules out residential proxy pools. How CloakBrowser compares: not published, so there is no comparable protected-site success rate to weigh against it.
#### AI agents that need to read blocked pages
Fortress ships a local MCP server with 29 agent tools, free, no account required. Claude, Codex, Cursor, Windsurf, and Cline can use it to read blocked pages, extract fields, and click, type, or fill forms. Setup instructions are in the repo's AGENTS.md. It also integrates with browser-use, crawl4ai, Stagehand, and LangChain. How CloakBrowser compares: no agent tooling or MCP integration is published for it.
#### CI and headless pipelines
Headless and headed modes present the same fingerprint, including on datacenter IPs, per the vendor. That means a scraper that passes in your dev environment does not silently fail when it moves to a headless CI runner. Each launch mints a distinct internally coherent machine, so parallel CI jobs do not share a fingerprint. How CloakBrowser compares: not published.
#### Teams that need license clarity
v3 is free for developers with no caps on machines, sessions, or concurrency. Activation is one command (tilion activate), the key is checked offline, and the engine never reports usage. Production use is free for individuals and smaller organizations, with funding and revenue thresholds; larger organizations need a subscription, with keys at fortress.tilion.com/pricing. Older engine code remains BSD-3. How CloakBrowser compares: its licensing terms are not published on any page we reviewed.
Best use cases for CloakBrowser
#### When you already run it in production
If your team has validated CloakBrowser against your specific target sites and it works, there is no reason to rip it out for a comparison page. Real-world results on your own site list beat any published benchmark, including Fortress's.
#### When its unpublished specs fit a niche you can verify yourself
Some anti-detect browsers target manual multi-account workflows with profile managers and GUI tooling rather than headless automation. If CloakBrowser serves that workflow for you and you have measured its pass rates on your targets, that is a legitimate reason to keep it. We cannot confirm or deny those capabilities from published material, which is itself the honest summary of this section.
Platform overview of Tilion Fortress
#### The engine
Fortress is a Chromium fork, currently v3 on Chromium 153.0.8010.36, updated monthly. Its 81+ C++ patches cover GPU, screen, fonts, timezone, and more, per the Fortress site. It is not a proxy service and not a hosted browser; it is the browser itself, running on your machines.
#### Installation and launch
Install via pip, npm, Docker, .deb, or a single binary, then launch and read the CDP endpoint:
``bash pip install tilion-fortress docker run --rm -p 9222:9222 tilion/fortress:latest ``
The Docker image exposes raw CDP on port 9222. Platforms shipped: Linux x64, Windows x64, Linux arm64/x86/armhf (no Widevine on ARM/x86 variants), and macOS arm64/x64 built from source.
#### Connecting from Playwright
```python from playwright.sync_api import sync_playwright
with sync_playwright() as p: browser = p.chromium.connect_over_cdp("http://localhost:9222") page = browser.pages[0] page.goto("https://tilion.dev") ```
No stealth plugins, no patch scripts, no extra flags. The npm package was still on Chromium 151.0.7910 at verification time; Node projects should connect over CDP to a separately installed v3 engine.
#### Activation and licensing
``bash tilion activate # saves key to ~/.tilion/license.jwt tilion license status --json TILION_LICENSE_KEY=<key> # for CI, Docker, and agents ``
Without activation the engine runs on the public v1 engine. The license check is offline and never reports usage. An environment key overrides the file, and an expired env key can shadow a fresh file, so check tilion license status if activation seems ignored.
#### MCP server for coding agents
Run tilion mcp to start the local MCP server: 29 tools, free, no account. Agents use it to read blocked pages, extract fields, and interact with forms. This is the fastest way to give Claude Code, Codex, Cursor, Windsurf, or Cline access to pages that currently return 403.
#### CLI surface
tilion [--port N] launches the engine and prints the CDP endpoint. tilion get [platform] downloads builds for linux-x64, arm64, armhf, x86, win-x64, mac-arm64, and mac-x64. tilion license refresh and tilion license logout round out license management. Note the Python wheel exposes tilion-fortress, and the module name is tillion_fortress with two l's.
How to choose decision matrix
| Your situation | Choose | |---|---| | You need published, reproducible stealth numbers before buying | Fortress (86.4% on 92 sites, harness in repo) | | You need a headless engine for CI with no fingerprint drift | Fortress (same fingerprint headless and headed) | | You want to keep existing Playwright/Puppeteer code unchanged | Fortress (one-line CDP connect) | | You need agent tooling (MCP, browser-use, crawl4ai, Stagehand) | Fortress (29-tool local MCP server) | | You need zero license cost as an individual or small team | Fortress (v3 free for developers, no caps) | | CloakBrowser already passes your target sites in testing | Stay on CloakBrowser; switching has no upside until you measure one | | You need a spec sheet or published pricing to evaluate an anti-detect browser | Fortress; CloakBrowser's specs are not published | | You need residential or mobile egress bundled today | Neither engine ships that; Fortress lists it on its roadmap, and Tilion Cloud (beta, waitlist at tilion.com) is the hosted option |
Migration guide: switching from CloakBrowser to Fortress
#### 1. Install the engine
Pick the channel that matches your deployment. For local testing, pip install tilion-fortress. For containers, pull the Docker image. For bare servers, tilion get linux-x64 fetches the binary.
#### 2. Point your automation at the CDP endpoint
CloakBrowser-style automation typically launches a patched browser binary and attaches to it. With Fortress you attach the same way, to a Chromium that is already correct:
```python
browser = p.chromium.launch(executable_path="/path/to/antidetect/chrome")
browser = p.chromium.connect_over_cdp("http://localhost:9222") ```
Delete your stealth plugin imports and fingerprint patch scripts. They are redundant against Fortress and can introduce their own detectable artifacts.
#### 3. Activate once
Run tilion activate locally, or set TILION_LICENSE_KEY in CI and Docker. Without activation you get the public v1 engine, which is not the engine the benchmark numbers describe, so do not evaluate v1 and conclude v3 failed.
#### 4. What has no equivalent
Fortress does not bundle proxies; the 86.4% benchmark ran on plain datacenter IPs. If your CloakBrowser setup depends on a bundled proxy network, you will need your own proxy provider alongside Fortress, or Tilion Cloud, which is in beta with a waitlist. Fortress also has no GUI profile manager for manual multi-account browsing; it is an engine for automation, not a desktop app. If that is your primary workflow, a partial migration (automation on Fortress, manual work elsewhere) is the better end state.
#### 5. Verify on your targets
Run your own 20-site list against Fortress before switching anything in production. The repo's docs/BENCHMARK.md harness is a reasonable template for a fair test: fresh machine per run, two reps, no proxies unless your production setup uses them.
FAQ
Is Fortress actually more effective than CloakBrowser at avoiding blocks?
Fortress loaded 86.4% of 92 protected sites in its own reproducible 28 September 2026 test, headless on datacenter IPs with no proxies. No comparable published figure exists for CloakBrowser, so a direct number-to-number comparison is not possible. Test both on your own target list.
Is Fortress free?
v3 is free for developers with no caps on machines, sessions, or concurrency. Production use is free for individuals and smaller organizations subject to funding and revenue thresholds; larger organizations need a subscription. Keys and terms are at fortress.tilion.com/pricing.
Do I need proxies with Fortress?
The headline benchmark did not use any: 86.4% of 92 sites on datacenter IPs alone. That said, site-specific results vary, and Fortress lists residential and mobile egress on its roadmap rather than shipping it today.
Does Fortress work with my existing Playwright code?
Yes, over CDP. chromium.connect_over_cdp("http://localhost:9222") is the only change. No stealth plugins or patch scripts, and removing them is recommended.
How does Fortress compare to Camoufox or puppeteer-extra stealth?
In the same 92-site run: Fortress 86.4%, Camoufox 74.5%, puppeteer-extra stealth 72.3%, stock Chrome 34.2%. All figures are from Tilion's own harness, methodology in the repo.
What is Tilion Cloud, and is it the same thing?
No. Fortress is the open-source engine you run yourself. Tilion Cloud is a hosted browser API built on that stealth work, currently in beta with a waitlist at tilion.com. ComputeSDK's 23 October 2026 run ranked Tilion Cloud first on both lifecycle (98.4) and throughput (90.9), and labels it "Pre-release".
If you are building scrapers or agents that keep getting blocked, download Fortress and run your own target list against it this week; the numbers above are reproducible, and so are yours.