LoginSign Up
← Back to all sparks
C

Calibre

ANALYTICS
Velocity6.3

Web performance monitoring with synthetic tests and real user monitoring

calibreapp.com ↗

Calibre MCP lets your coding agent read RUM and Lighthouse data, then fix the slow code

web-performancemcpcoding-agentsrumlighthousecore-web-vitals
◆Current state
Calibre now ships a read-only MCP server that gives Claude Code, Cursor, Codex, Copilot, Claude and ChatGPT access to an account's RUM, CrUX and Synthetic data, included on every plan. It came days after Test Agents moved to Chrome 156 and Lighthouse 13.6.0. The latest entry is a design tidy-up that places each page's Guide link in the same top-right spot.
◆Where it's heading
Calibre is turning its monitoring data into something agents act on rather than something people read in dashboards. The summer work on deployment markers, all-pages aggregates and the audit log built the context and governance an agent needs: which deploy, which page, who connected. Expect the remaining surface to follow the same path, with agents both querying performance and being measured by it.
◆Prediction
The likely next step is write access or deploy-linked workflows over MCP, such as an agent opening a fix tied to a regression marker, since the current server is explicitly read-only and the launch frames fixing as the goal.

◆Recent moves

  1. 4d ago

    Deploys: see which release made your site slower

    Calibre's RUM and Synthetic monitoring now compare each release with the one before it, shade regressed deploys in red and show which metrics and pages got worse. It ties performance data to the release that caused it, a useful step for a monitoring tool but not a change of direction.

    View source ↗
  2. 5d ago

    Guides: everything in its right place

    Guide links now sit at the top right of every page that has one. A small consistency fix after the MCP launch, not a change in what the product measures.

    View source ↗
  3. 11d ago

    Calibre MCP: connect your coding agent to your performance data

    ⚡ SPARK

    The MCP server is where the year's groundwork pays off: deployment markers, all-pages aggregates and audit logging now feed an agent instead of a chart. It moves Calibre from a dashboard you check to a data source your coding agent queries before it edits.

    View source ↗
  4. 14d ago

    Chrome 156 & Lighthouse 13.6.0

    Test Agents move to Chrome 156 and Lighthouse 13.6.0, whose LCP and FCP fixes will lift many Performance Scores on PageSpeed profiles without any site change. Snapshots also add Agentic Browsing audits, which Google marks as under development.

    View source ↗
  5. 1mo ago

    Core Web Vitals Test, now with ad metrics

    The free Core Web Vitals Test adds four experimental ad metrics from CrUX on the day Google announced them. It is a free-tool change rather than a core product one, but shows how quickly new browser data reaches Calibre.

    View source ↗
  6. 2mo ago

    Synthetic Metrics: All pages, segmented by Test Profile and Deployment

    Synthetic Pulse and metrics show median aggregates across all pages instead of one page at a time, with segmented views by test profile, deployment, and page. The by-page segment is the useful one — it answers whether a regression is site-wide or confined to a single route.

    View source ↗