89%lighter page
ShakaCode cut the Snorkel.com customizer page weight by 89% and hit 100% Lighthouse accessibility, shipped in a 10-day final-mile launch with zero checkout regressions.
Read the Snorkel storyFree report
Start with a free report on your site
Send a URL. In about 15 minutes you get a private, unlisted link to a full report: what mobile visitors, screen readers, and AI crawlers actually experience on your pages.
- Most AI crawlers, such as OpenAI, Anthropic, and Perplexity, fetch your HTML but do not execute JavaScript. A page can rank on Google and still look near-empty to AI answer engines. The report shows exactly what they get.
- Every report runs the full pipeline, never a metered preview. When a metric is already good, the report says so instead of inventing work.
- Reports are private by default: unguessable link, noindex, sent only to you.
Mobile speedNeeds work
AccessibilityNothing to fix here
AI visibilityNear-empty to AI crawlers
Three status tiles open every report, followed by an annotated load timeline and prioritized fixes.
Proof, not promises
This workflow ships measured wins
ShakaPerf is the toolkit ShakaCode uses in client engagements, where every claim has to survive a production launch.
Since 2018performance partner
ShakaCode has run performance and infrastructure for Popmenu, a SaaS platform serving thousands of restaurant locations, holding Core Web Vitals steady as the product keeps growing.
Read the Popmenu storyHow it works
How ShakaPerf proves a gain
You would not time two sprinters on different days, one with a headwind, and call it a fair race. ShakaPerf runs both branches at the same instant on the same machine, so the noise cancels and the difference is real.
Containerize
Wrap your app in a production Docker image with one exposed port and write a single Playwright test. That one test drives every check that follows.
A/B test
Control and experiment branches run simultaneously in separate containers, sampled together so shared CPU noise cancels, then compared with a paired Wilcoxon test. Trustworthy p-values without a quiet machine or a dedicated CI box.
Prove
Only changes that move the numbers ship. Visual, accessibility, and SEO checks run in the same pass, so speed never costs you correctness.
Guard
Once ShakaPerf is in CI, it watches your main branch and flags performance regressions automatically, before they reach production.
Want the statistics in full? The tests, the exact p-value computation, and the published thresholds are documented in used_statistics.md.
Capabilities
One toolkit, every check that guards a release
One test, every check
Write one Playwright test and get performance benchmarks, visual regression, accessibility, and network-activity checks from the same definition.
Statistically rigorous A/B testing
Paired Wilcoxon analysis with real p-values and CPU-noise cancellation, so results hold up without a dedicated CI box.
Visual regression detection
Screenshot comparison between main and feature branches on multiple screen sizes, so an optimization never silently breaks the UI.
Stack agnostic
If it ships in a Docker container, ShakaPerf measures it. Rails, Node, Python, Go, no rewrite required.
Accessibility and SEO tests
WCAG 2.0 to 2.2 at levels A and AA, plus best practices. Every optimization stays SEO-compliant.
Automatic regression detection
No gradual degradation from entropy and rapid development. Your main branch stays fast.
Built for coding agents
The performance signal your agent is missing
Unit tests tell an agent whether behavior broke. Type checks tell it whether the code compiles. Nothing tells it whether the app got faster, until now.
- An agent proposes an optimization, ShakaPerf proves or refutes it with real p-values, and the PR merges with quantified evidence attached.
- Reports are readable by humans and machines, so agents can act on results inside CI loops.
- The docs and license are written to be read by agents too. Point your agent at shakaperf.com and let it wire things up.
agent · session
# in your agent session
> use ShakaPerf to verify this PR
# what the agent runs
yarn add shaka-perf
shaka-perf compare
significant
# what comes back
LCP -54% (p < 0.01) · visuals unchanged
bundle -12% · a11y 0 new issues
Quickstart
About two hours to your first proven result
If you already have a production Dockerfile, setup is a config file and a Playwright test away. Runs locally and in CI.
- A production Docker image with one exposed port
- One Playwright test that exercises the page you care about
- Node 20+, Docker, Overmind, and GNU parallel on the machine
# install yarn add shaka-perf shaka-bundle-size # compare your branch against main shaka-perf compare # full report: perf, visual, a11y, bundle shaka-perf audit
Source-available, enforced by trust
No license keys, no activation, nothing phones home. Evaluation (45 days per organization, agents welcome), education, personal projects, and public open-source work are free for everyone, and production use is free for small organizations, including CI and coding-agent workflows.
Larger organizations subscribe. Registration is free, optional, and encouraged for every organization. See pricing and the license.
Use it yourself, or let us run it
Read the docs and self-host ShakaPerf, or request a free report and the team that builds it will show you what your visitors actually experience.