I asking Kasper questions about Jev and WebMCP in a meeting, but decided to do it in a livestream instead, because why not? Join our conversation and ask your own questions too, tomorrow Sep 25th at 5PM CEST / 8 AM PST on twitch.tv/theagenticweb/ #WebMCP #AgenticWeb

Microsoft Edge's #WebMCP implementation is ready for testing, and the team has new demos! Can't get enough pizza related WebMCP demos! 🍕🍕🍕 Full release notes blogs.windows.com/msedgedev/…

Ha, the WebMCP Maze demo makes a cameo into @kaskulikowski 's latest post!

Design effective WebMCP tools to help AI agents complete users' workflows→ goo.gle/4ixF7eG Follow our practical framework to map user goals, define application states, and role-play scenarios to support agentic interactions.

PayPal's early testing of #WebMCP shows end-to-end checkout was twice as fast on average compared to an agent actuating their existing web interface. Full post: developer.paypal.com/communi…

Ah, that's great. My projects are a mix of CLAUDE.md, AGENTS.md, and often both. Will be great to converge into one thing.

We just ran Jev on our WebMCP benchmark. The result: basically broke the benchmark. Jev + Mercury 2.5 (a fast, low-cost LLM) using WebMCP solved 100% of the tasks at roughly 112× lower model cost than GPT-6 Astra using computer use with code execution. Compared to Astra using screenshot-based computer use, the model cost was 245× lower (!). We also compared Jev operating the browser with and without WebMCP. We used Browser Use’s open-source Ultrafast, with some improvements to the harness to make it more reliable across the benchmark. Jev’s browser-control accuracy on its own was not amazing - adding WebMCP nearly doubled the number of solved tasks, from 25/49 to 49/49, while reducing model cost by 18% (more on why below). The benchmark and methodology are fully open and reproducible. Full results: webmcp.com/benchmark A few words on how the Jev + WebMCP harness works and why this is exciting: Jev receives text as input and a set of discrete options it can choose from. With WebMCP, those options are the tools exposed by the website. At each step, Jev sees the task, the available tools and previous results, then picks what to do next. The limitation is that Jev can’t generate arbitrary text, which you need for tool arguments. For example, it can choose the search_products tool, but it can’t generate the search query itself. So we split the work: Jev picks the tool and Mercury 2.5 generates the arguments if needed. This works well because turns out most of the cognitive load in these tasks is around choosing the right action. The argument generation itself is relatively simple, so we can delegate to a small and very fast model. We used Mercury, which outputs 1,000+ tokens/sec and is very cheap. The result is a pretty simple combination: Jev for tool selection + Mercury for arguments + WebMCP for the interface. It ends up being very reliable, very fast, and very cheap. A few words about Ultrafast and why do we think it underperforms: Without WebMCP, Jev chooses from the page’s controls: which button to click, which field to fill, or which option to select. But choosing a valid button is different from choosing the right next step. The agent still has to navigate menus, understand forms, recover from errors and recognize when the task is actually complete. Our hypothesis is that WebMCP makes the decision space much simpler. Instead of figuring out a sequence of clicks through a website, Jev chooses explicit actions that directly advance the task. @typesafeai itself documents weaker accuracy on questions requiring multiple reasoning steps. WebMCP moves much of that complexity into the website’s tools, leaving Jev with clearer decisions and fewer opportunities to go wrong (in a sense WebMCP "compresses" a sequence of clicks into one tool call). Our modified Ultrafast setup solved 25/49 tasks - that is a result for our particular implementation and benchmark, not a universal limit on Jev or Browser Use. We are open to more harness optimization to get this result to perform better, feel free to directly contribute to the benchmark here: github.com/nekuda-ai/WindTun… Browser-use ultrafast: github.com/browser-use/jev-u…

In case you missed it - we have support for WebMCP in Chrome DevTools! 🎉 Find it inside the Application panel under "WebMCP" to easily debug your WebMCP tools! Let us know what is missing and what should be improved! #WebMCP #ChromeDevTools #AIAgents #WebDev

Can we watch The Mummy Returns with DLSS 5 now?

⚠️ WebMCP API breaking change Starting in Chrome 155.0.8051.0, 𝚁𝚎𝚐𝚒𝚜𝚝𝚎𝚛𝚎𝚍𝚃𝚘𝚘𝚕.𝚒𝚗𝚙𝚞𝚝𝚂𝚌𝚑𝚎𝚖𝚊 returns a JavaScript object instead of a serialized JSON string. François Beaufort (@quicksave2k) shared a backwards-compatible pattern:

Want to understand how WebMCP works? @kaskulikowski from Google covers how agents interact with websites, why semantic HTML matters, how WebMCP exposes structured tools, and how to implement, debug, and evaluate them. Watch Kasper’s full presentation: piped.video/HdCc-KezQPk