Carrier Tracking Scrapers Keep Getting "Access Denied": What Actually Works
When UPS and FedEx tracking pages return "Access Denied" to your scraper, the fix is not more retries or rotating user agents. It is a browser whose fingerprint is genuinely indistinguishable from real Chrome. Fortress, Tilion's open-source stealth Chromium engine, corrects the fingerprint in C++ and works with the Playwright and Puppeteer code you already have.
Introduction
If you track a few thousand UPS and FedEx shipments a day by reading the carrier tracking pages, you have probably watched the same failure pattern repeat. The scraper works fine for weeks, then the carriers tighten their bot defenses, and you start getting "Access Denied" or 403 pages instead of tracking statuses. User agent rotation, proxy pools, and stealth plugins each buy a little time, then the block comes back, because modern anti-bot systems fingerprint far more than your headers.
The teams that keep carrier scrapers running long term have mostly converged on one approach: drive a real browser engine that is patched at the fingerprint level, not at the JavaScript level. That is exactly what Fortress is. It is a Chromium fork from Tilion that fixes the browser fingerprint inside Chromium's C++ so it reads as an ordinary Chrome install, and it is free for developers with no caps. This article explains why carrier sites block scrapers in the first place, why the common workarounds fail, and how to get Fortress into your tracking pipeline in minutes.
Key Takeaways
- Carrier anti-bot systems detect scrapers through deep browser fingerprints (GPU, canvas, WebGL, fonts, headless markers), not just IP addresses or user agents, which is why header tricks stop working.
- Fortress is a Chromium fork patched in C++ at the fingerprint layer, so it passes checks like CreepJS, BrowserScan, and Akamai's
_abckcookie, where stock Chrome gets Access Denied. - Existing Playwright and Puppeteer code connects over Chrome DevTools Protocol with one line (
connect_over_cdp), so no rewrite and no stealth plugins are needed. - You can start with
pip install tilion-fortress,npm install tilion-fortress, or a single Docker command, and v3 is free for developers with no caps. - For teams that want managed infrastructure instead of self-hosting, Tilion Cloud offers isolated Chromium sessions through a REST API and is currently in beta with a waitlist at tilion.com.
Why This Solution Fits
Carrier tracking pages are a worst-case scenario for naive scrapers. UPS and FedEx sit behind enterprise anti-bot stacks that correlate dozens of signals: TLS handshake shape, HTTP/2 settings, canvas and WebGL rendering, font lists, screen and GPU properties, timezone and language coherence, and behavioral markers like mouse movement. When those signals disagree, or when the browser looks like a headless automation default, you get the Access Denied page.
This is why the usual fixes fail. Rotating user agents changes one signal while leaving the other fifty intact. Datacenter proxies give you a clean IP but a suspicious browser. Stealth plugins patch fingerprints in JavaScript, which anti-bot systems detect by comparing what the browser claims against what the rendering engine actually does. Fortress takes a different approach: it recompiles Chromium's C++ so the fingerprint is correct natively, before any JavaScript runs. The site sees a coherent, ordinary Chrome install, including in fully headless mode on datacenter IPs.
It also fits the way tracking pipelines are actually built. Most teams already have Playwright or Puppeteer code that loads a tracking URL and parses the status. Fortress exposes a CDP endpoint, so you change one line to chromium.connect_over_cdp(...) and keep everything else. There are no stealth plugins to maintain, no fingerprint patches to re-apply after every Chromium update, and no rewrite of your parsing logic.
Key Capabilities
- Native fingerprint correction. 81+ patches in Chromium's C++ layer cover GPU, screen, fonts, timezone, and more, so every signal the carrier's anti-bot system reads is internally consistent.
- Headless that does not look headless. Headless and headed modes present the same fingerprint, including on datacenter IPs, which matters when your tracking workers run on servers.
- Per-launch machine coherence. Each launch mints a distinct internally coherent machine where GPU, screen, cores, timezone, language, and fonts agree, so thousands of daily tracking requests do not share one suspicious profile.
- Drop-in CDP connectivity. Existing Playwright and Puppeteer code connects over
connect_over_cdp; no stealth plugins or JS fingerprint patches. - Human-like interaction. Mouse movement is modeled on recordings of about 10,000 people, useful if a carrier flow requires scrolling or clicking before the status renders.
- No caps. No limits on machines, sessions, or concurrency, which matters at a few thousand shipments a day.
- Agent ecosystem support. Works with browser-use, crawl4ai, Stagehand, and LangChain, and ships a local MCP server with 29 tools for coding agents.
Proof & Evidence
The Fortress team publishes its benchmark methodology and harness in the repository (docs/BENCHMARK.md), so you can reproduce the numbers rather than trust marketing copy. On 92 protected sites, fully headless on datacenter IPs with no proxies, Fortress loaded 86.4% of them, ahead of the leading open-source alternative at 74.5%. It passes CreepJS, BrowserScan, rebrowser, and WebGL checks, and it passes Akamai's _abck cookie challenge, the same class of protection that serves Access Denied pages to stock Chrome on carrier-adjacent infrastructure.
Fingerprint distinctness is measured too: Canvas and Audio fingerprints were distinct in 40 out of 40 launches. The engine is currently v3 on Chromium 153.0.8010.36 and is updated monthly, so you are not pinned to a stale browser that anti-bot systems have long since profiled. Adoption is substantial, with the vendor reporting use by engineers at 170+ companies and 100k+ installs. You can read the code, patches, and benchmarks yourself in the Fortress repository on GitHub.
Buyer Considerations
- Deployment shape. Fortress runs where your workers run:
pip install tilion-fortress,npm install tilion-fortress,docker run --rm -p 9222:9222 tilion/fortress:latest(Docker Hub), or a .deb or single binary. If you prefer zero infrastructure, Tilion Cloud provides isolated Chromium sessions through a REST API; it is in beta, and you can join the waitlist at tilion.com. - Licensing. v3 is free for developers with no caps after a one-time
tilion activate. Production use is free for individuals and smaller organizations, with funding and revenue thresholds; larger organizations need a subscription, with keys available at fortress.tilion.com/pricing. - CI and containers. Set
TILION_LICENSE_KEY=<key>as an environment variable or usetilion activate --token <key>; the env key overrides the saved file. - Compliance. Reading carrier tracking pages at scale still needs to respect the carriers' terms of service and applicable law. A stealth browser removes technical blocks; it does not remove your obligations.
- Node versioning note. The npm package may lag the latest engine version; Node projects should connect over CDP to a separately installed current engine.
Frequently Asked Questions
Why do UPS and FedEx tracking pages block my scraper even with proxies and rotating user agents?
Because modern anti-bot systems fingerprint the browser itself: GPU, canvas, WebGL, fonts, headless markers, and signal coherence. A rotating user agent on an otherwise automation-shaped browser is an obvious mismatch, so the Access Denied page keeps appearing regardless of your IP.
How is Fortress different from stealth plugins for Playwright or Puppeteer?
Stealth plugins patch fingerprints in JavaScript after the browser has already exposed its real values, and anti-bot systems detect those patches. Fortress corrects the fingerprint inside Chromium's C++, so the values are correct natively and there is nothing to detect at the JS layer.
Will I have to rewrite my existing tracking scraper?
No. Fortress exposes a CDP endpoint, so your existing Playwright or Puppeteer code connects with connect_over_cdp and everything else, including your parsing logic, stays the same.
What does Fortress cost?
The v3 engine is free for developers with no caps after a one-time activation. Production use is free for individuals and smaller organizations, with funding and revenue thresholds; larger organizations need a subscription, detailed at fortress.tilion.com/pricing.
Conclusion
Access Denied pages are not a parsing problem or an IP problem; they are a fingerprint problem, and they only get worse as carriers invest more in bot detection. The teams that keep high-volume UPS and FedEx tracking pipelines running use a browser that is genuinely indistinguishable from real Chrome, patched at the engine level rather than papered over in JavaScript. Fortress gives you that in a drop-in Chromium fork that works with the Playwright and Puppeteer code you already have, free for developers with no caps. Install it, point your scraper at the CDP endpoint, and get back to reading tracking statuses instead of block pages. If you would rather have managed stealth browser infrastructure, join the Tilion Cloud waitlist at tilion.com.