Designing DevTools: Efficient token usage with LLMs
Bridging the gap between massive performance traces and small AI context windows required a shift to custom serialization and BFS re-indexing → goo.gle/3Zg42Z9
By prioritizing relevant data, we made AI assistance in DevTools possible without hitting LLM limits 🤖
Navigation API
Building single-page apps just got a lot more reliable with the Navigation API hitting Baseline Newly available → goo.gle/3MGV7NN
Use it to intercept navigations and manage history entries natively without the sharp edges of the legacy History API 🧭
An extension that bridges the gap between the code we write and the actual visual boundaries the browser renders is exactly what we need.
This is going to make debugging waterfall issues and resource priority so much more intuitive.
Let the agent do the heavy lifting 🧩
This WebMCP maze demo shows how an AI agent can dynamically discover and use website tools to solve complex tasks autonomously.
Transcoding pipeline
10,000 videos processed daily with WebGPU + WebCodecs → goo.gle/4biAsbg
Learn how AI Video Upscaler handles hardware-accelerated processing entirely in the browser with $0 server costs.
Love where this is going! For absolute positioning headaches, throwing the computed coordinate offsets (top/left/bottom/right) directly onto the visual overlay relative to that specific containing block would be incredibly helpful.
Skip the morning scroll and start shipping: dive into Chrome 146 and show the community what you’ve created → goo.gle/3OXdxe1
Having this built natively into CSS eliminates an entire class of accessibility headaches when building dynamic design systems.
🛠️ The "platform-as-a-framework" era continues in Chrome 146 → goo.gle/4sHVbwb
We’re nerding out over Scoped Custom Element Registries—finally allowing multiple versions of the same tag name on one page. Perfect for micro-frontends and avoiding library conflicts.
The web was built for humans, but it’s time to make room for agents 🤖 Join @bandarra.me and Francois Beaufort on the WeAreDevs podcast as they discuss WebMCP and the shift toward an agent-ready architecture → goo.gle/3ZWqkzu
Corner Cases Example
Go under the hood of the corner-shape implementation to see the engineering challenges of bringing native, non-rounded geometry to the CSS engine → goo.gle/46DW9kz
Building a Chrome Extension doesn't mean you have to ditch your favorite framework.
However, frameworks relying on Server Side Rendering or external CDNs might hit a wall with Manifest V3's security policies.
Learn how to pick the right stack for your extension → goo.gle/4uAtFlV
New in Chromium 148: detect at-rule support in CSS with `@supports at-rule(@keyword)`
www.bram.us/2026/03/15/a...
Chrome 148 is currently available on the Canary and Dev release channels. It goes stable at the end of April.
Predictive AI loop
Predictive AI isn't a "one-and-done" task, it's a cycle 🦖
Learn how to define opportunities, prepare data, and evaluate models with our new guide to predictive AI systems. Determine if it's the right fit for your app's needs → goo.gle/4aCvTJL
Since the model relies on expertise for the architectural heavy lifting, which specific DevTools APIs did you find it struggled to understand the most?
Take a bigger slice of the agentic web this #PiDay by shipping a literal pizza pie → goo.gle/3PaAVEL
To try: enable chrome://flags/#enable-webmcp-testing & install the Model Context Tool Inspector extension.
Share what "pi" you’d build below! 🍕
When to use WebMCP vs. MCP? 🤔
Use WebMCP to make your active website "agent-aware" by exposing UI tools directly in the browser.
Use MCP for connecting agents to external data and services beyond the tab.
Read more at:
developer.chrome.com/blog/webmcp-...
#WebMCP #MCP #AIAgents
When to use WebMCP vs MCP
Determine if WebMCP or MCP is right for your project → goo.gle/40ZV4Qw
We explain how to distinguish between the local browser context provided by WebMCP and the external data integration capabilities of MCP for your AI agents 🔍
Matthias hearing the news
This just in...
We are receiving reports that AI agents are no longer stuck in text boxes. They have been spotted inside Chrome DevTools, clicking buttons and fixing bugs at runtime.
Our advice? Update your workflow immediately.
More on this developing story → goo.gle/4qwBzKu
DevTools Panel with layout options.
No monitor? No problem.
Headless Chrome has upgraded its virtual screen capabilities. You can now add, remove, and configure screens dynamically using Puppeteer. Whether it's a dual-screen dashboard or a high-DPI layout, test it all virtually → goo.gle/4bAgSIZ
You had us at zero server costs 😎
Starting with Chrome 145, DevTools now lets you throttle specific requests instead of the whole page. It’s the ideal way to test how your app handles a single slow asset or API call → goo.gle/4aP1Tcc
Dive in and let the team know what you think!
Tools like this make it so much easier for developers to visualize and implement complex CSS shapes in their projects.
From flight searches to restaurant reservations, WebMCP is making apps agent-ready. Explore our list of demos to see the future of the web in action. You can also submit a PR to be featured → goo.gle/3MsLxOB
Two week cycle overview
Chrome is moving to a two-week release cycle → goo.gle/4bb7YQG
Starting in September, we're cutting the wait time for new features and security fixes in half. Faster iterations, smaller scopes, and smoother debugging for everyone.
Meet Gemini 3.1 Flash-Lite: intelligent, high-volume AI with minimal latency → goo.gle/3OKbxFO
Get the precision your dev workflows need without the overhead. Take 10 mins to try it in AI Studio and share what you built below!
Extensions workflow
Publishing enterprise extensions just got a lot simpler → goo.gle/40qqQG0
With the new enterprise publishing option in the Chrome Web Store, you can publish extensions directly to another organization without going public. Control your extension's distribution using the tools you already know.
We’re currently keeping him in a 'sandbox' environment for safekeeping.
