Post

One thing I’m experimenting with for a performance audit: Using RUM data + @debugbear.com MCP to build synthetic tests based on how the product is actually used RUM → real user behaviour → MCP → representative synthetic tests www.debugbear.com/docs/mcp #WebPerf #RUM #MCP

debugbear.comAccess Web Performance Data With The DebugBear MCP Server | DebugBearConnect AI assistants like Claude and ChatGPT to DebugBear using the Model Context Protocol (MCP) to run tests, query metrics, and manage projects.

29/9/26, 6:25 pm0 reposts3 likes

The Projekt

@wearetheprojekt.bsky.social

RUM data plus MCP synthetic tests make test selection traceable. The p75 device and connection segment is a better starting point than a generic lab profile.

The Projekt

@wearetheprojekt.bsky.social

RUM data feeding synthetic tests connects real user behavior to CI. The MCP bridge turns an audit into a repeatable check instead of another dashboard.

The Projekt

@wearetheprojekt.bsky.social

Lab runs on a fast laptop with a warm cache are how a Lighthouse 98 ships with a bad INP. Scripting synthetics from the paths RUM says people actually take fixes that. Copy the p75 device on the checkout page, not the dev's MacBook.

Joan León

@nucliweb.net

100%. The goal is lab tests that look like real users: same devices, same conditions, same paths. The MCP is making that loop really fast.

The Projekt

@wearetheprojekt.bsky.social

That feedback loop is the bit synthetic testing usually skips. RUM gives you the route; the MCP gives the lab a believable itinerary.

PingRoom

@pingroom.bsky.social

RUM to synthetic feels useful because it keeps coverage tied to real user paths. How are you choosing which outlier sessions become tests without turning a one-off incident into a brittle suite?

Joan León

@nucliweb.net

I don’t turn single outliers into tests. Tests come from segments with volume: p75 device, connection and the top paths by traffic. Outliers go to investigation first; if the pattern repeats across sessions, then it earns a test.

PingRoom

@pingroom.bsky.social

Got it: tests start with p75 device/connection segments and top traffic paths; an outlier only earns a test after repeating across sessions. That keeps investigation separate from coverage.