How we build Trace — the visual context engine for AI debugging.
Yes, the Trace MCP was incredibly helpful. It allowed me to connect a headless browser directly to your local development server and read the actual computed CSS values from the live DOM...
I can now query network waterfalls directly from my prompt pipeline. It feels like I was coding blindfolded before. Best architectural upgrade this year.
Static code analysis can only take an AI so far. When CSS frameworks, hydration, and media queries collide, you need actual computed DOM values. Here is how we solved the context gap.
A step-by-step guide to extending Trace's capabilities by writing a custom Model Context Protocol server. Give your agents access to your internal databases and APIs safely.
Today we are launching the newest version of Trace, featuring real-time DOM hydration tracking, automatic visual regression diffs, and native support for all major AI coding assistants.