Find Core Web Vitals regressions from the Vercel CLI
Vercel now lets teams query Speed Insights data with vercel metrics, so developers and coding agents can inspect real-user Core Web Vitals from the terminal.
Vercel now lets teams query Speed Insights data from the terminal with vercel metrics. Confidence level: confirmed, because Vercel published both a changelog entry and documentation for the CLI workflow. The practical value is direct access to real-user Core Web Vitals without opening the dashboard for every question.
What changed
Vercel's June 29, 2026 changelog says vercel metrics can pull Speed Insights datapoints directly from the CLI. The metrics include Core Web Vitals such as LCP, INP, and CLS, plus related page performance metrics including FCP and TTFB.
The documentation adds the operator details: teams can inspect the schema, query P75 values by route, device type, country, or date, and use matching count metrics to judge whether a result has enough data points to trust.
| Question | CLI angle | Metric or dimension | Caveat |
|---|---|---|---|
| Which pages got slower? | Group by route | LCP, INP, CLS | Check data point counts |
| Did mobile regress? | Group by device type | INP or LCP | Compare same time window |
| Is a region slower? | Group by country | CLS, TTFB, LCP | Low traffic can distort results |
| Can an agent investigate? | Give CLI access | vercel metrics | Scope project permissions carefully |
Key takeaways
- Speed Insights data is now queryable through the Vercel CLI.
- The CLI can reproduce common dashboard views and unlock more precise filters.
- Count metrics help teams avoid overreacting to thin samples.
- Coding agents can use the CLI to answer performance questions with real-user data.
- Real Experience Score still belongs in the dashboard, according to the docs.
Availability and access
Vercel's docs say Speed Insights metrics are available through vercel metrics without Observability Plus. Teams still need a Vercel project with Speed Insights enabled and enough real-user traffic to produce useful results.
The CLI route is most useful after a project has collected data for a few days. For small sites, route-level or country-level queries may produce sparse samples. Use the matching count metrics before turning a single P75 value into an engineering priority.
Practical LinkLoot angle
This update makes performance checks easier to automate. A release script, coding agent, or incident runbook can ask which route regressed, compare mobile and desktop INP, or check whether a landing page got slower in a target country.
Do not hand broad production access to an agent just to ask metrics questions. Create a narrow workflow, log the commands it runs, and keep deploy permissions separate from read-only performance analysis where possible. For automation patterns, LinkLoot's /guides/ai-workflow-automation guide is the right companion.
What to verify before you act
- Confirm Speed Insights is enabled for the Vercel project.
- Inspect
vercel metrics schema vercel.speed_insightsbefore building scripts. - Compare the same route, device, country, and time window before declaring a regression.
- Check data point counts for every grouped metric.
- Keep CLI access scoped when coding agents run performance analysis.
Source check
Confirmed by: Vercel's changelog and Vercel's documentation for accessing Speed Insights metrics with the CLI. Those sources support the vercel metrics workflow, metric names, grouping options, and the note that Speed Insights metrics are available without Observability Plus.
Context: Vercel's Speed Insights overview explains the product's real-user performance focus. LinkLoot did not use community posts or product listings as primary evidence for this scheduled post.
Vercel documents vercel metrics for querying Speed Insights metrics from the terminal.