Non-Coder's Guide to OpenClaw (2): fast-browser-use - The Reality of "10x Faster Browser Automation" 🤦‍♂️

Introduction

“Ultra-fast browser automation built with Rust! 10x faster than Puppeteer!”

…and how I got baited into 2 hours of struggle 🤦‍♂️

🎣 Taking the Bait

I discovered a skill called fast-browser-use on ClawHub. The description promised:

Rust-based: Startup time < 50ms (Puppeteer takes ~800ms)
Memory: 15 MB (Puppeteer uses 100+ MB)
Bot Detection Bypass: Human emulation features
Infinite Scroll Harvesting: Auto-collect SNS feeds

“This is exactly what I’ve been looking for!” 🤩

What really caught my eye was the “Bypass Bot Detection via Human Emulation” feature. Access even tough sites like Coupang? (Coupang is Korea’s Amazon - the largest e-commerce platform here)

🛠️ The Installation Journey

Step 1: Installing the Skill

clawhub install fast-browser-use

✅ Installation complete! (That was too easy… ominous)

Step 2: Rust Build

cd workspace/skills/fast-browser-use
cargo build --release

Build time: 1 minute 9 seconds

Downloaded and compiled 330 packages. Something feels suspiciously heavy…

Result: 18MB binary (Wait, wasn’t it supposed to be 15MB memory? 🤔)

Step 3: Chrome Installation

WSL doesn’t have Linux Chrome by default. Starting from scratch!

wget https://dl.google.com/linux/direct/google-chrome-stable_current_amd64.deb
sudo dpkg -i google-chrome-stable_current_amd64.deb

Result:

✅ Setting up google-chrome-stable (xxx.xxxx.xxx-x)...

Oh! That was easy? (Again, ominous…)

Step 3.5: First Run… and Dependency Hell 😱

Confidently trying to run it:

google-chrome-stable --version

Error bombardment:

google-chrome-stable: error while loading shared libraries: 
libnspr4.so: cannot open shared object file: No such file or directory

😭 “Missing library?!”

Turns out Chrome needs system libraries that aren’t installed. This is what happens when running GUI apps on WSL…

Attempt 1: Install basic libraries

sudo apt install -y libnss3 libnspr4

Try again:

google-chrome-stable --version

Another error:

error while loading shared libraries: libatk-1.0.so.0: cannot open shared object file

😤 Fix one, another appears!

Attempt 2: Install all Chrome dependencies

Googled the full list of Chrome requirements:

sudo apt install -y \
  libnss3 \
  libnspr4 \
  libatk1.0-0 \
  libatk-bridge2.0-0 \
  libcups2 \
  libdrm2 \
  libdbus-1-3 \
  libxkbcommon0 \
  libxcomposite1 \
  libxdamage1 \
  libxfixes3 \
  libxrandr2 \
  libgbm1 \
  libpango-1.0-0 \
  libcairo2 \
  libasound2 \
  fonts-liberation

Yet another error:

Note, selecting 'libatk1.0-0t64' instead of 'libatk1.0-0'
Note, selecting 'libatk-bridge2.0-0t64' instead of 'libatk-bridge2.0-0'
Note, selecting 'libcups2t64' instead of 'libcups2'
Package libasound2 is a virtual package provided by:
  liboss4-salsa-asound2 4.2-build2020-1ubuntu3.1
  libasound2t64 1.2.11-1ubuntu0.1
You should explicitly select one to install.
E: Package 'libasound2' has no installation candidate

😱 “Package names changed?!”

Turns out Ubuntu 24.04 upgraded some libraries to t64 versions with different names.

Attempt 3: Reinstall with corrected package names

sudo apt install -y \
  libnss3 \
  libnspr4 \
  libatk1.0-0t64 \
  libatk-bridge2.0-0t64 \
  libcups2t64 \
  libdrm2 \
  libdbus-1-3 \
  libxkbcommon0 \
  libxcomposite1 \
  libxdamage1 \
  libxfixes3 \
  libxrandr2 \
  libgbm1 \
  libpango-1.0-0 \
  libcairo2 \
  libasound2t64 \
  fonts-liberation

Result:

✅ 17 packages installed

Try again:

google-chrome-stable --version

Result:

Google Chrome xxx.0.xxxx.xxx

🎉 Finally success!!!

💡 Lesson: Ubuntu 24.04 users beware of t64 versions!

Starting from Ubuntu 24.04, time_t changed from 32-bit to 64-bit, causing some library names to change:

libatk1.0-0 → libatk1.0-0t64
libcups2 → libcups2t64
libasound2 → libasound2t64

WSL users, remember this!

🎯 First Test: Coupang Search

Now let’s search for “Monster Energy Drink” on Coupang (Korea’s Amazon)!

fast-browser-use screenshot \
  --url "https://www.coupang.com/np/search?q=몬스터%20에너지" \
  --output coupang_monster.png \
  --full-page

Result:

✅ Saved screenshot to "coupang_monster.png"

Oh! Did it work?!

Opening the screenshot…

Access Denied

You don't have permission to access "http://www.coupang.com/np/search?" on this server.

😭 Detected as a bot

🕵️ Problem Analysis

Checking the User-Agent:

Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/xxx.xx (KHTML, like Gecko) 
HeadlessChrome/xxx.0.0.0 Safari/xxx.xx

Problems:

HeadlessChrome ← Literally advertising “I’m a bot”
X11; Linux x86_64 ← Environment rarely used by normal users

Coupang: “Lol, it’s a bot? Blocked~”

💡 Solution Attempt 1: Human Emulation

Re-reading SKILL.md:

fast-browser-use navigate --url "https://protected-site.com" \
  --human-emulation \
  --wait-for-selector "#content"

Oh! There’s a --human-emulation option!

fast-browser-use navigate \
  --url "https://www.coupang.com/np/search?q=몬스터%20에너지" \
  --human-emulation \
  --wait-for-selector "body"

Result:

✅ Navigation complete.

Success…?

But the problem:

navigate command only navigates to pages
Need screenshot command to capture
screenshot has no --human-emulation option 😱

Trying SKILL.md’s “Login & Cookie Heist” method:

# Step 1: Open real browser for manual login
fast-browser-use login --url "https://www.coupang.com" --save-session coupang.json
# (Browser opens → Manual login → Press Enter)

# Step 2: Screenshot with saved session
fast-browser-use screenshot \
  --url "CoupangURL" \
  --load-session coupang.json \
  --output result.png

Result:

error: unexpected argument '--load-session' found

😭 screenshot command doesn’t support --load-session!

The Truth Revealed

Session loading support by command:

✅ navigate: Supported
❌ screenshot: Not supported
❌ markdown: Not supported
❌ snapshot: Not supported

You can save sessions, but can’t actually use them! 🤦‍♂️

🍎 The Real Reason: Mac vs WSL

Found a success case on YouTube:

Title: “Coupang Search Success”
Environment: Mac

The differences:

	Mac	WSL (My Environment)
OS	macOS	Linux
User-Agent	`Macintosh; Intel Mac OS X`	`X11; Linux x86_64`
Headless Detection	More natural	`HeadlessChrome` explicit
GUI	Native	WSLg (Virtual)

Conclusion: Works on Mac, blocked on WSL/Linux 😭

📊 Final Comparison: Brave Search API vs fast-browser-use

Feature	Brave Search API	fast-browser-use
Speed	⚡ ~1s	🐌 ~5s
Stability	✅ 100%	⚠️ ~30% (Many sites block)
Environment Dependency	❌ None	⚠️ Mac » WSL/Linux
Bot Detection	✅ None	❌ Often blocked
JavaScript Rendering	❌ No	✅ Yes
Screenshots	❌ No	✅ Yes (if not blocked)
Complex Interactions	❌ No	✅ Yes
Coupang Access (WSL)	✅ Via API	❌ Blocked

🎓 Lessons Learned

1. Don’t trust “10x faster” marketing

Startup time: 50ms? → Actual total operation: 5+ seconds
Memory: 15MB? → Binary alone is 18MB

2. Environment matters

What works on Mac might not work on Linux
Bot detection checks down to OS level

3. Check CLI tool completeness

Having a feature doesn’t mean it works (session saving works but unusable)
Inconsistency between commands makes tools useless

4. Don’t dismiss existing tools

Brave Search API: Boring but always works
fast-browser-use: Looks cool but environment dependent

🗑️ Ending: Deletion

After 2 hours of struggle, my choice:

rm -rf fast-browser-use
sudo apt remove google-chrome-stable

Returned to: Brave Search API

“Stable tools beat shiny new ones” - Post-struggle wisdom

If you use browsing a lot, fast-browser-use would be decent choice

💬 Conclusion

fast-browser-use itself isn’t a bad tool. However:

When it’s useful:

✅ Mac environment
✅ Sites with weak bot detection (GitHub, Hacker News)
✅ When JavaScript rendering is essential
✅ When screenshots are absolutely needed

When to avoid:

❌ WSL/Linux environment
❌ Sites with strong bot protection (Coupang, Naver Shopping)
❌ When you need fast, reliable results
❌ General web searches

Lesson: Struggles are learning experiences. But next time, check environment compatibility first 🤦‍♂️

P.S. This post was co-written with my AI assistant. He/She summarized our 2-hour struggle in 3 minutes. AIs are good at documenting their own frustrations too lol

Introduction#

🎣 Taking the Bait#

🛠️ The Installation Journey#

Step 1: Installing the Skill#

Step 2: Rust Build#

Step 3: Chrome Installation#

Step 3.5: First Run… and Dependency Hell 😱#

💡 Lesson: Ubuntu 24.04 users beware of t64 versions!#

🎯 First Test: Coupang Search#

🕵️ Problem Analysis#

💡 Solution Attempt 1: Human Emulation#

💡 Solution Attempt 2: Login & Session#

The Truth Revealed#

🍎 The Real Reason: Mac vs WSL#

📊 Final Comparison: Brave Search API vs fast-browser-use#

🎓 Lessons Learned#

1. Don’t trust “10x faster” marketing#

2. Environment matters#

3. Check CLI tool completeness#

4. Don’t dismiss existing tools#

🗑️ Ending: Deletion#

💬 Conclusion#

References#

Introduction

🎣 Taking the Bait

🛠️ The Installation Journey

Step 1: Installing the Skill

Step 2: Rust Build

Step 3: Chrome Installation

Step 3.5: First Run… and Dependency Hell 😱

💡 Lesson: Ubuntu 24.04 users beware of t64 versions!

🎯 First Test: Coupang Search

🕵️ Problem Analysis

💡 Solution Attempt 1: Human Emulation

💡 Solution Attempt 2: Login & Session

The Truth Revealed

🍎 The Real Reason: Mac vs WSL

📊 Final Comparison: Brave Search API vs fast-browser-use

🎓 Lessons Learned

1. Don’t trust “10x faster” marketing

2. Environment matters

3. Check CLI tool completeness

4. Don’t dismiss existing tools

🗑️ Ending: Deletion

💬 Conclusion

References