@ulpi/browse

How browse compares

Feature-by-feature comparison with the major headless browser tools for AI agents.

Feature@ulpi/browseagent-browserbrowser-use
Navigation
Navigate to URL
Back / Forward / Reload
Get current URL
Content Extraction
Page text extraction
Element HTML
Links extraction
Forms extraction
Accessibility tree
Interaction
Click / Fill / Select
Hover / Focus / Drag
Keyboard input
File upload
Wait conditions (element, URL, network)
Snapshot & Refs
@ref element selection
Interactive-only snapshot
Viewport-scoped snapshot
Cursor-interactive detection
Snapshot diff
Visual
Screenshot (viewport / full / element)
PDF export
Responsive screenshots
Visual diff
Annotated screenshots
Sessions & Tabs
Tab management
Session multiplexing
Persistent profiles
iframe targeting
Recording
HAR recording
Video recording
Command recording & export
Network mocking
React DevTools
Component tree
Props / State inspection
Suspense / Error boundaries
Hydration timing
Render profiling
Security
Domain allowlist
Encrypted credential vault
Content boundaries (prompt injection defense)
Agent Integration
Handoff protocol (CAPTCHA/MFA)
Claude Code skill
JSON output mode
Device emulation (100+ devices)
Total Features15213651

Benchmarks

Token usage (10-action session)

@ulpi/browse11,388 tokens
browser-use27,331 tokens
agent-browser31,886 tokens

Speed (multi-step flow)

@ulpi/browse28.5s
browser-use72.7s

Benchmarks from standardized 10-action browsing sessions. Lower is better.