Skip to main content

The ultimate web performance toolkit

Pulsates runs real Lighthouse audits and exposes every metric, asset, and timing to your AI coding agent. From cryptic Core Web Vitals scores to merged PRs — in one session.

Get a Pro plan to configure audits for your needs
Get a Pro plan to configure audits for your needs
today · 10:42
Can you check the latest audit for example.com and tell me the biggest issues?
Claude · via MCP

Your LCP is 4.2s (poor) and TBT is 890ms (poor). The LCP element is the hero image — 1.4MB with no preload hint.

TBT is dominated by a 1.2MB stylesheet loaded render-blocking in <head>, uncached on first visit.

Which stylesheet is it and where does it come from in our codebase?
Claude · via MCP

Blocking stylesheet /assets/main-Bq3k2.css is compiled from src/styles/main.scss.

I see 847 @extend calls in _base.scss — many nested inside media queries, which stops SCSS from deduplicating selectors and bloats the output.

Supercharge your audits with MCP
Wire production metrics into Claude, Cursor, or Copilot
Install MCP$ npx @pulsates/mcp
MCP connected · 11 tools·claude-sonnet-4.5
⌘K · commands
Every response grounded in your own audit data · nothing hallucinated

Features

Built for the full performance cycle

Pricing

Start free. Scale when it matters.

Hobby
Free

Personal projects and exploration.

  • 10 audits / month
  • Mobile & desktop
  • Core Web Vitals
  • Diagnostics report
  • 7-day history
Get started
Pro
$19/mo

Unlimited audits and full MCP access.

  • Unlimited audits
  • MCP server · 11 tools
  • Asset source traces
  • Chrome UX Report data
  • Deploy webhooks
  • 90-day history
Start free trial
Team
$49/mo

Shared performance ownership across your team.

  • Everything in Pro
  • 10 projects · 5 seats
  • Audit comparison diffs
  • Slack & webhook alerts
  • Priority support
Start free trial
14-day money-back guarantee·No credit card to start·Cancel anytime

FAQ

Common questions

PageSpeed Insights shows you a score card in a Google-controlled environment. Pulsates returns every metric, asset, timing, and diagnostic as structured JSON your AI agent can query directly. You also control device, network throttle, CPU slowdown, and test region — so you can reproduce exactly what slow users experience.

Run `npx @pulsates/mcp` once and point it at your API key. Claude, Cursor, and Copilot immediately get 11 tools they can call to query audits, look up specific metrics, trace assets to their source file, and check CrUX field data — without leaving the editor.

Any tool that supports the Model Context Protocol. That includes Claude (claude.ai and API), Cursor, GitHub Copilot, and any custom agent you build with an MCP-compatible runtime. The server exposes a standard stdio transport.

Most audits complete in 3–5 minutes. We run a full Lighthouse pass in a headless browser throttled to your specified conditions. You receive a webhook notification when results are ready, and your agent can query them immediately.

No — CrUX requires a site to have sufficient real-world traffic in Chrome's data collection. For low-traffic or brand-new URLs we return lab data only and note that field data isn't available yet.

Yes. We expose a webhook endpoint you can call from any pipeline — GitHub Actions, GitLab CI, Vercel deploy hooks, Netlify build hooks. The audit runs automatically and your agent can verify improvements before the PR merges.

7 days on the Hobby plan. 90 days on Pro and Team. Historical data powers the before/after diff view and lets your agent spot regressions across deploys.

Not yet. Pulsates currently audits publicly accessible URLs. Authentication support — cookie injection and form-based login flows — is on the roadmap.

Want to know more?

Take a look at our documentation