Every page's vitals. Ranked by what to fix first.
Vitalsly puts New Relic performance, Sentry errors and Lighthouse scores on one row per page, then tells you which page is costing you the most and why.
$ npx @karan171996/metricflowFrom keys to a ranked list of pages in about five minutes.
Connect your tools
Paste a New Relic key, a Sentry token, or both. You only see the tools you connect. Lighthouse needs no key.
NEW_RELIC_API_KEY=NRAK-•••••• SENTRY_AUTH_TOKEN=sntrys_••••••
Add one call to your site
No vendor SDK to wire up. The Connect page writes this snippet with your own values filled in.
import { init } from '@karan171996/metricflow/browser'; init({ sentry: { dsn }, 'new-relic': { … } });
Fix the worst page first
Every page is marked Healthy, Warning or Critical by limits you set, with an AI note on the likely cause.
● /checkout Critical LCP 3.8s ● /search Warning INP 310ms ● /blog Healthy LCP 1.2s
One row per page. Every signal on it.
The merged per-page view
Load time, LCP, TTFB and CLS from New Relic, error counts from Sentry and Lighthouse scores, side by side.
Honest numbers
A value nobody measured shows as a dash, never a misleading zero.
Your thresholds, your trends
Set warning and critical limits once. Hosted plans keep 30 to 90 days of history and week-over-week changes.
AI that names a cause
Ask Claude, Gemini or OpenAI what to fix first. Bring your own key locally, or use ours on hosted plans.
Alerts where your team is
In-app notifications and Slack when a page crosses a limit. Hosted plans only.
| Page | Status | LCP | Errors | LH |
|---|---|---|---|---|
| /checkout | Critical | 3.8s | 142 | 54 |
| /search | Warning | 2.9s | 37 | 71 |
| /product/[id] | Warning | 2.6s | 12 | 78 |
| / | Healthy | 1.6s | 4 | 92 |
| /blog/launch | Healthy | 1.2s | — | 97 |
Free on your machine. Hosted when it matters in production.
The npm package is MIT licensed, with no account, no licence check and no server of ours. Hosted adds what needs a server: history, alerts, teams and scheduled Lighthouse runs.
npx, and you're running
Runs on localhost with your own keys. Built for checking your site while you develop it.
- All three tools: New Relic, Sentry, and Lighthouse CI against your dev server
- Keys stay in a file on your machine; nothing is sent to us
- AI suggestions with your own Claude, Gemini or OpenAI key
- Top 200 pages, one project per running instance
- Latest 24 hours only, no stored history
- One person, no Slack alerts, development data only
Production, staging and your team
Sign in with GitHub or Google. Your site pushes snapshots to your workspace, and we store them and do the work.
- Development, staging and production environments
- 30 to 90 days of history with trends and week-over-week changes
- Scheduled Lighthouse runs on your staging and production URLs
- In-app and Slack alerts when a page crosses a limit
- AI suggestions included, no key needed
- Invite teammates as owner, admin or member
- Ask us for a copy of your data, or to delete it, at any time
Free for localhost. Paid for production.
Every plan gets every tool. Plans differ in where they run, how much history they keep and how many people use them.
For one developer checking a site on localhost.
- All three tools
- Your own AI key
- Latest 24 h
- 1 seat · 1 project
- GitHub issues support
For a solo developer with a site in production.
- 3 projects · 50 pages each
- 30 days of history
- 300 Lighthouse runs / month
- Slack alerts · 1 channel
- 100 AI insights / month
- Email support
For a product team sharing one view of performance.
- 5 seats, then $12 / seat
- 10 projects · 200 pages each
- 90 days of history
- 2,000 Lighthouse runs / month
- Slack alerts · 3 channels
- Priority email support
For larger teams that need longer history and closer support.
- 13+ months of history
- Custom webhooks
- Audit log and export (planned)
- Optional EU region (planned)
- Support SLA (planned)
Every limit side by side is on the full plan comparison.
Built to read your data, not keep it.
Calls to us from local mode
The npm package talks only to New Relic, Sentry and the AI provider you pick. It never calls a Vitalsly server.
Hosted never holds your vendor keys
Your site pushes snapshots to your workspace with an ingest token you can revoke at any time.
Open source local package
Read the code that runs on your machine. The hosted service is a separate codebase.
Questions people ask first.
Is the local version really free?
Yes. It’s an MIT-licensed npm package with no account and no licence check. Run it on as many machines as you like.
What’s the difference between Vitalsly and MetricFlow?
MetricFlow is the free npm package you run locally. Vitalsly is the hosted service built on it, with history, alerts and teams.
Do I need both New Relic and Sentry?
No. Connect either one, or both. Lighthouse works with no account at all. You only see the tools you connect.
Why is production data a paid feature?
Our costs come from storing history, sending alerts, calling the AI and running Chrome for Lighthouse. The paid plans cover those. Local mode does none of them on our side, so it stays free.
When does the beta end, and what happens to my data?
Beta workspaces get notice before paid plans start. To get a copy of your workspace data or have it deleted, send us a message through the support form and we’ll help.
Which AI providers do you support?
Claude, Gemini and OpenAI. Locally you bring your own key. Hosted plans include a monthly number of insights on our key.
Stuck? We answer.
GitHub issues
Bugs and questions about the free local package.
Email support
Help with your hosted workspace and setup. Send us a message.
Troubleshooting guide
What to check when the dashboard is empty or a number looks wrong.
Getting started
The keys you need and how to check that metrics arrive.