More Amplifier Stories
🌐

Browser Automation
with Amplifier

"When fetch isn't enough"

Active
The Problem

The Gap

What traditional tools can't do:

🚫
No JS Execution
SPAs render empty divs. React, Vue, Angular - all invisible.
🔒
No Authentication
Can't fill login forms. Gated content stays gated.
👁️
No Visual Verification
Can't "see" the page. Layout bugs go unnoticed.
🖱️
No Interaction
Can't click, type, or navigate. Static reads only.
Modern web apps need a real browser.
The Solution

Browser Bundle

What amplifier-bundle-browser provides:

🤖
browser-operator
Navigate, click, fill forms
📸
visual-documenter
Screenshot & document
🔍
web-researcher
Extract & analyze data
Key Capabilities
  • Natural language control - just describe what you want
  • ~93% more efficient - 700 tokens vs 8,000 (CLI-based)
  • Ref-based selection - stable targeting (@e1, @e2)
  • 8 ready recipes - quick demos to SPA testing
Case Study

Amplifier SDK Development

The Meta Story: Amplifier Building Itself

Context
Example web app with TypeScript frontend + Python server. Goal: Self-validate the SDK during development.
What browser automation enabled:
  • Launch the example app in a real browser
  • Interact with SDK features as a real user would
  • Validate that TypeScript bindings work correctly
  • Test the Python server integration end-to-end
  • Report issues found during automated exploration
Key Insight
"Amplifier used browser automation to develop and validate its own SDK"
Case Study

Workspaces UI Validation

User Journey Testing at Scale

Context
Workspaces web UI with complex project creation flows. Goal: Validate journeys, find gaps before users do.
What browser automation enabled:
  • Walk through complete user journeys end-to-end
  • Create projects through the UI (not API shortcuts)
  • Test what real users would actually experience
  • Report missing features found during exploration
  • Document gaps with screenshots as evidence
Key Insight
"Found missing features by being a user, not reading code"
Bonus

Data Extraction

Reach dynamic content that fetch can't touch

🕸️
Dynamic SPAs
Content loaded via JavaScript, infinite scroll, lazy loading - all accessible.
# What you get with fetch: <div id="app"></div> # What you get with browser: <div id="app"> <table>...actual data...</table> </div>
📊
Real-World Uses
  • Extract data from internal dashboards
  • Scrape paginated results with "Load More"
  • Capture data behind authentication walls
  • Export from apps without API access

Shows versatility beyond testing - agents can gather data from anywhere

Philosophy

Be a User,
Not a Script

🔍

Explore Naturally

Scroll, click, be curious. Don't follow rigid scripts - discover like a user would.

📸

Strategic Screenshots

Only capture when meaningful. Document issues, not every page load.

🤝

Ask When Stuck

Human-in-the-loop for ambiguous situations. Collaboration over automation.

🧠

Think Like the End User

What would confuse them? What would delight them? Test for experience.

Sources

Research Methodology

Data as of: February 20, 2026

Feature status: Active (v0.3.0)

Repository: amplifier-bundle-browser

Research performed:

  • Git log analysis: git log --oneline (10 commits, all Jan 30, 2026)
  • Contributor analysis: git log --format="%an" (1 contributor)
  • README review: token efficiency claim (93% reduction)
  • Directory scan: find . -name "*.md" for agents and recipes

Gaps: Recipe count (8) and agent count (3) from documentation claims, not verified in repo file structure. Token savings figure (93%) from README, not independently benchmarked. Case studies are based on usage narratives, not measured outcomes.

Primary contributor: Sam Schillace — 10 commits (100%). Built in a single day.

Get Started

Try It Now

Bundle
samueljklee/amplifier-bundle-browser
Quick Demo
examples/quick-demo.yaml
SPA Testing
recipes/spa-ui-test.yaml
3
Agent Patterns
8
Recipes (per docs)
~93%
Token Savings

Give your agents eyes and hands.

More Amplifier Stories More Amplifier Stories