sitespeed.io measured from a real browser: what the report holds
sitespeed.io is an open-source tool for comprehensive web performance analysis, enabling you to test, monitor, and optimize your website’s speed using real browsers in various environments.
At a glance
- What is it?
- sitespeed.io loads your page in Firefox, Chrome, Edge or Safari and collects Core Web Vitals, video, a HAR waterfall and Coach advice. The install path is short, and almost every knob lives on the documentation website instead of in the README.
- Who is it for?
- sitespeed.io fits a team that wants one browser measurement it can keep, file into a pull request and ship to a time series database, and it fits best when someone owns the browsers it drives. It fits badly if you need field data from real users next to lab numbers, a documented budget file format, or Safari measurements from a container.
- Can I use it commercially?
- Yes. MIT is a permissive licence: you can use, modify and sell software built on it, as long as you keep its copyright and licence notices.
- Is it still maintained?
- Yes. The repository received new commits within the last day.
- What is it written in?
- Mainly JavaScript, according to GitHub's language statistics.
Answers come from the project's GitHub data, last synced on November 15, 2026, and from our analysis. They are not legal advice.
Star history
- sitespeedio/sitespeed.io5,053 ★ · +23
| Repository | 2026-09-23 | 2026-11-15 |
|---|---|---|
| sitespeedio/sitespeed.io | 5k | 5.1k |
23 Stars in the last 53 days, measured from the daily GitHub snapshot. Compare with other projects
Editorial analysis
Docker carries the browsers, npm leaves them to you
The Docker route is called the easiest way to run the tool because the image ships Firefox, Chrome, Edge and the dependencies, and one command mounts the current directory so results land where you can find them:
docker run --rm -v "$(pwd)":/sitespeed.io sitespeedio/sitespeed.io https://www.sitespeed.io/The npm route is two steps, and the second one is where the work lands:
npm i -g sitespeed.iositespeed.io https://www.example.comAfter that you need a browser installed locally, Firefox, Chrome, Edge or Safari, and FFmpeg plus Python if you want video and visual metrics. That is the whole difference between the two routes: the container carries the browsers and their dependencies, while the global install assumes the machine already has them. The shorter path depends on whether you are already running containers or already running a browser toolchain.
The container installs a published ADB keypair and a sudoers rule
The image is built on `sitespeedio/webbrowsers:chrome-156.0-firefox-159.0-edge-156.0-20261101`, sets two browsertime environment variables, `SITESPEED_IO_BROWSERTIME__XVFB` and `SITESPEED_IO_BROWSERTIME__DOCKER`, and points `PYTHON` at `python3`. WebPageReplay is installed at build time from a certificate pair that is copied into the image:
RUN wpr installroot --https_cert_file /webpagereplay/certs/wpr_cert.pem --https_key_file /webpagereplay/certs/wpr_key.pemAndroid support is why two details in the file deserve attention. The Dockerfile adds `docker/adb/insecure_shared_adbkey` as the root ADB key, with the comment that the point is to avoid clicking the OK button on a device. It also allows all users to run the commands needed by sitespeedio/throttle via sudo, with a comment saying only the subcommands throttle uses are allowed.
Anyone attaching a real phone to that container is trusting a key pair committed to the repository, and rebuilding the image with your own key is the only way to avoid it.
Five iterations and a median, because one run is noisy
The measurement advice is a single flag. Add `-n 5` and the tool runs five iterations and reports the median, which the project calls recommended for any real measurement because single runs are noisy:
sitespeed.io https://www.example.com --browser chrome -n 5The browser is selected with a long flag or a short one, `--browser chrome` here and `-b safari` further down, and everything else is behind `--help`:
sitespeed.io --helpTwo things follow for a team writing a budget check. First, a CI gate built on one iteration inherits that noise, so a regression rule that fires on the median of five is a different rule from one that fires on the first run. Second, the default iteration count is not written down here, so the number a run uses when you pass no `-n` at all has to be checked against `--help` or the configuration documentation rather than assumed to be one.
Real phones arrive over USB, and Safari is not in the image
Two mobile paths exist and neither one is emulated. Real Android phones are driven with:
sitespeed.io https://www.example.com --androidReal iPhones are driven with Safari, and the README marks that path as new in 40.0:
sitespeed.io https://www.example.com -b safari --safari.iosBoth are documented as phones connected over USB, with separate setup guides for Android and iOS. The practical limit is the container: the published Docker image carries Chrome, Firefox and Edge, and Safari on iOS is not part of it, so that measurement needs hardware and a host outside the image. Nothing here describes a device farm, a cloud device service, or a way to choose between several attached models, so a lab with a mixed set of phones needs a decision made outside this project.
The report holds five kinds of evidence and no field data
What comes out of a run is fixed. Core Web Vitals, namely LCP, INP, CLS, TTFB and FCP, scored against Google's p75 thresholds. Visual metrics, namely First Visual Change, Speed Index and Last Visual Change, plus a scrubable filmstrip, captured with `--video --visualMetrics`. The HAR waterfall, rendered with waterfall-tools. The Coach's advice, a checklist of best practice rules with severity, scores and concrete fixes. And CPU and long task analysis of what the main thread was doing during the load.
Every one of those is a lab measurement taken by one browser on one machine. The README documents no way to compare them against real user measurements, no RUM integration and no field correlation, so a score that disagrees with what users report has to be reconciled by hand. The budget mechanism that would turn these numbers into a gate is also not spelled out here; it appears only as a use case.
Budgets, scheduling and metric shipping point back at the website
Three ways to use the tool are named, and all three end in a link. Audit a page from your terminal, run on every pull request and fail the build if a budget is exceeded, or run on a schedule, ship the metrics to Graphite or InfluxDB and watch them over time in Grafana. A live instance of that last setup is published at `dashboard.sitespeed.io`.
What the README does not contain is the file format for a budget, the property that names a destination, or a sample configuration file. Budgets are a documented use case with no documented artefact, and the metric destinations are named as products rather than as settings. The repository does show where report rendering lives, since the lint target of the project is `lib/plugins/html/templates`, so report templates are part of the tree even though the configuration is not.
Version 42 arrives on a shrinkwrapped dependency tree
`package.json` carries version 42.10.0, matching the v42.10.0 release published on 2026-10-23, after v42.9.0 on 2026-10-15 and v42.8.0 on 2026-10-13. The last push to `main` was on 2026-11-01 and the repository is not archived. Dependencies are held by `npm-shrinkwrap.json`, which is the file the Dockerfile copies next to `package.json` before running `npm ci --omit=dev`, so an image build follows the pinned tree rather than a range.
The package installs two binaries, `sitespeed.io` pointing at `./bin/sitespeed.js` and `sitespeed.io-wpr` pointing at `./bin/browsertimeWebPageReplay.js`. Scripts cover `ava` for tests, `eslint. && npm run pug-lint` for linting and `tools/check-licenses.js` for licence checks. The documentation site carries a second toolchain of its own, with a `Gemfile`, `netlify.toml` and a `.spelling` list. A project that has shipped since 2012 on a version in the forties is moving often, and no upgrade notes beyond `CHANGELOG.md` and `ROADMAP.md` sit in the checkout.
Editorial conclusion
sitespeed.io fits a team that wants one browser measurement it can keep, file into a pull request and ship to a time series database, and it fits best when someone owns the browsers it drives. It fits badly if you need field data from real users next to lab numbers, a documented budget file format, or Safari measurements from a container. Before adopting, confirm FFmpeg and Python are present if video and visual metrics matter, check the iteration count your budget will run with, and read the configuration documentation before writing your first budget rule.
Frequently asked questions
what is sitespeed io
sitespeed.io is an open source web performance tool that drives a real browser, Firefox, Chrome, Edge or Safari including Safari on a real iPhone over USB, to load a page. It then collects Core Web Vitals, video and visual metrics, the HAR waterfall, the Coach's advice and CPU and long task analysis.
How do I install sitespeed.io from npm?
Run `npm i -g sitespeed.io`, then `sitespeed.io https://www.example.com`. You need a browser installed locally (Firefox, Chrome, Edge or Safari), plus FFmpeg and Python if you want video and visual metrics. The Docker image instead ships Firefox, Chrome and Edge with their dependencies.
Why should sitespeed.io run more than one iteration?
Adding `-n 5` runs five iterations and reports the median, which the project recommends for any real measurement because single runs are noisy. The README does not state how many iterations a run uses when no `-n` flag is given.
Can sitespeed.io test a page on a real iPhone?
Yes. `sitespeed.io https://www.example.com -b safari --safari.ios` runs Safari on a real iPhone over USB, a path the README marks as new in 40.0. Real Android phones use `--android`. The published Docker image carries Chrome, Firefox and Edge only, so the Safari run needs real hardware.
Official sources
Add this badge to your README
If you maintain this project, the badge below links readers to this analysis and shows its maintenance status from the daily GitHub snapshot. Paste the markdown into your README; add ?metric=license or ?metric=stars to the image URL for a different field.
[](https://hysenlabs.com/projects/sitespeedio-sitespeed-io)Star history chart for your README
[](https://hysenlabs.com/projects/sitespeedio-sitespeed-io)Was this useful?