Core/Dash vs SpeedCurve

What SpeedCurve does well

If you work in web performance you've probably learned a thing or two from SpeedCurve. Their blog and their web performance guide are some of the best free reading on the subject, and people from SpeedCurve helped shape the standards Core Web Vitals are built on. The BBC, The Guardian, GOV.UK and Forbes all use it.

speedcurve.com, October 2026.

SpeedCurve has real user monitoring and synthetic testing in one place, and that combination is still the main reason to buy it. You get scheduled lab tests in Chrome and Firefox, scripted user journeys, Lighthouse, and filmstrips and comparison videos. Those videos make a slow page obvious to people who never open DevTools. Try explaining render delay to a marketing director without one.

The RUM part is good too. SpeedCurve shows LCP subparts, the LCP element, INP subparts and the element someone interacted with. Since May 2025 it reports Long Animation Frames with the script that caused them, split into first and third party and filterable by INP phase. It reports LCP and INP from Firefox and Safari as well as Chrome, has a proper SPA mode for soft navigations, and can tie performance to conversion rate and bounce rate. Then there are budgets, alerts to Slack, email and Microsoft Teams, a GitHub and CircleCI integration, competitor benchmarking, a daily raw export to S3 and unlimited users on every plan.

First, measure your real users

If you use SpeedCurve today and it answers your questions, keep it. You're ahead of most of the sites I look at.

The ones I worry about have no real user monitoring at all. They look at a Lighthouse score (Lighthouse never clicks anything), or at CrUX. CrUX only counts Chrome users who opted in, lumps 28 days of traffic into one number and shows nothing when a site is too small (see not enough data in Search Console). SpeedCurve, Core/Dash or any other decent RUM tool fixes that, so just pick one and install it.

The rest of this page is for people who are still choosing, or who keep running into questions their current tool can't answer.

Core/Dash and SpeedCurve side by side

Pricing and features checked September 2026.

Why I'd still pick Core/Dash

AI agents can read your data

For me this is the big one. Core/Dash has a built-in MCP server at https://app.coredash.app/api/mcp. Add it as a connector in Claude, or plug it into Claude Code, Cursor, Windsurf, VS Code or Gemini CLI, and your agent can query the field data itself. Ask which pages have the worst INP on mobile this week and if it's getting worse. It calls get_metrics grouped by page and filtered to mobile, then get_timeseries for the trend, and you get an answer with your own numbers. Access is read-only, the agent can't change anything. MCP is included in every Core/Dash plan, Starter too.

I also built CWV Superpowers, a free, MIT-licensed agent that runs on top of it. It takes a problem Core/Dash found in the field, reproduces it in a Chrome trace and comes back with a fix (file, line and diff) or a report. There's a write-up on fixing INP with an AI agent if you want to see it in action.

I couldn't find an MCP server or any AI features in SpeedCurve itself. Embrace, the parent company, has an MCP server, but its docs only cover Embrace's own data and don't mention SpeedCurve or Core Web Vitals. More and more diagnosis work is done by agents, and an agent can only use data it has access to.

Same story for scripts. SpeedCurve's RUM query API is in beta and only returns data for a chart you've already built in the UI, 200 buckets max. The Core/Dash API is the same endpoint the agents use: filter or group by any dimension and get up to 500 segments back. For agencies there's an agency API as well, to create and manage client projects from your own tooling.

Pricing and how long you keep the data

At 500K pageviews a month Core/Dash Standard is $31 billed annually. SpeedCurve starts at $90 for the same traffic, 3K synthetic checks included. Past that SpeedCurve charges per pageview, and their own docs suggest sampling to keep costs down. Sampling is fine for a sitewide trend line. But ask for the p75 INP of one checkout template on Android in Germany and you run out of data quickly, and those narrow segments are often where the problem is.

Then there's history. SpeedCurve plans under $500 a month keep 3 months of RUM data. Core/Dash Standard keeps a year, Studio 3 years and Enterprise 8. Was INP this bad last Black Friday? Did the redesign from two summers ago help? You can't check that with three months.

SpeedCurve support docs on data retention, October 2026.

A bit more attribution detail

SpeedCurve's attribution is good, this is about details. Their docs say LCP subparts only apply to images and video. If your LCP element is text you see dashes. On a lot of content sites the LCP element is a heading or a paragraph, and then it's nearly always render delay, so that's a real gap. Core/Dash splits every LCP into the four phases and adds the element type, plus whether the image was preloaded, high priority or lazy loaded.

For INP Core/Dash records the input type and the load state. A slow click while the page is still loading needs a different fix than a slow click on a page that finished loading ten seconds ago, and it helps a lot to know which one you're dealing with. The Diagnosis view names the vendor script (GTM, HubSpot, Intercom) and gives fixes for your stack: WordPress, Shopify, Next.js and eight others. Your CrUX data is in the same dashboard as your RUM data, so you can see what Google sees next to what your visitors get.

Only Core Web Vitals

Core/Dash only does Core Web Vitals RUM and everything I build goes into that. SpeedCurve has been part of Embrace since November 2025, and in May 2026 Embrace launched its own web RUM and synthetic product, built on SpeedCurve's technology. That means more engineers, and both companies say nothing changes for SpeedCurve customers. I believe them. Still, if you're choosing a tool for the next five years it's fair to wonder where the combined roadmap is going.

When to pick SpeedCurve instead

• You want lab and field data in one tool, with scheduled synthetic tests, scripted journeys and filmstrips. Core/Dash only has a pre-deploy check, there's no synthetic suite.
• You look at competitor benchmarks every week and management wants the videos.
• Your company runs on Microsoft Teams, needs private synthetic agents or wants SpeedCurve's consulting team.
• You've got years of budgets and dashboards in SpeedCurve that the whole company uses. If it answers your questions today, I'd leave it.

Try Core/Dash next to what you have

You don't have to decide on day one. The Core/Dash script is under 2KB and loads async, so you can install it next to SpeedCurve, run both during the 14-day trial and compare the numbers. No credit card. See pricing (or agency pricing if you manage client sites), or book a demo and I'll go through your own data with you. Looking at other tools as well? All comparisons are on the compare page.

Sources: SpeedCurve pricing, SpeedCurve data retention, SpeedCurve Core Web Vitals subparts, SpeedCurve LoAF support, SpeedCurve sample rate, SpeedCurve Queries API, SpeedCurve joins Embrace, Embrace web RUM launch, Embrace MCP docs.