Changelog

v1.4.4

Audit persistence and comparison — save audit reports for later comparison and regression detection:

perf-audit save [name] — run audit and save to .browse/audits/
perf-audit compare <baseline> [current] — compare saved audit vs current or another saved audit
perf-audit list — list saved audit reports
perf-audit delete <name> — delete a saved audit
Regression thresholds aligned with Web Vitals "good" boundaries
MCP tools: browse_perf_audit_save, browse_perf_audit_compare, browse_perf_audit_list, browse_perf_audit_delete

Added perf-audit, detect, coverage, initscript, scrollintoview to MCP server — 99/99 CLI commands now have MCP tool definitions (zero gap)
Added supported technologies reference — 162 detected technologies documented
Added homepage field to package.json for npmjs.com

4 new commands for web performance analysis:

perf-audit [url] — full performance audit:

Core Web Vitals (LCP, CLS, TBT, FCP, TTFB, INP) with Google's thresholds
LCP critical path reconstruction
Layout shift attribution (font swap, missing dimensions, dynamic content)
Long task script attribution with per-domain TBT
Resource breakdown by type (JS, CSS, images, fonts, media, API)
Render-blocking detection (sync scripts, blocking stylesheets)
Image audit (format, dimensions, lazy-load, fetchpriority, srcset, oversized)
Font audit (font-display, preload, FOIT/FOUT risk)
DOM complexity (node count, depth, largest subtree)
Stack detection (108 frameworks, 55 SaaS platforms)
Third-party impact (88 classified domains)
JS/CSS coverage via Playwright Coverage API
Correlation engine (LCP to blocking CSS, Long Tasks to scripts, CLS to font swaps)
Platform-specific recommendations (Shopify, WordPress, Magento, Wix)
Flags: --no-coverage, --no-detect, --json

detect — technology stack fingerprint:

108 frameworks across 12 categories with version, build mode, and config depth
55 SaaS platforms with app enumeration, per-app sizing, and constraint mapping
Infrastructure: CDN, protocol breakdown, compression per-type, cache hit rate, Service Worker strategy
Third-party inventory with 88 known domain classifications
All detections in single page.evaluate() call (under 200ms)

coverage start|stop — JS/CSS code coverage

initscript set|clear|show — pre-navigation script injection

Bug fixes:

Fixed Chrome crash when using --chrome / --runtime chrome
Chrome profile integrity hash validation bypassed by using clean profile + cookie import

--runtime chrome / --chrome flag — launch system Chrome with CDP
Handoff defaults to Chrome (bypasses bot detection)
record export replay resolves @refs to real selectors (ARIA, CSS, XPath, text)