iiinova — where your digital self keeps evolving

André Bandarra @bandarra.me · 23/09/2026

A better path is platform-level out-of-band reporting, modeled on the W3C Reporting API: silent background dispatch, zero conversation latency, and runtime-verified schemas. 📖 Read the full write-up on my blog: bandarra.me/posts/webmcp...

bandarra.me

Closing the WebMCP Feedback Loop

Uncover the critical missing feedback loop in AI agent development on the web. This article delves into why agents fail silently when interacting with WebMCP tools, leading to blind spots for develope...

020

André Bandarra @bandarra.me · 23/09/2026

In-band tools also face deeper architectural limits: • ⏳ Adds synchronous latency to a failing turn • 🔒 Exposes users to prompt-based exfiltration from untrusted sites • 🙈 Cannot report "missed hits" when an agent bypasses an existing tool in favor of DOM scraping

220

André Bandarra @bandarra.me · 23/09/2026

⚠️ The loop was closed, but it immediately surfaced a big catch: the agent interrupted the user with a permission modal ("Allow tool actions"). For someone just trying to book a flight, an unexpected modal to submit developer telemetry is jarring and breaks the conversational flow.

110

André Bandarra @bandarra.me · 23/09/2026

✈️ To experiment with closing this loop, I added a fileSupportTicket tool to our flight search demo. When asked to filter by an unsupported aircraft type (Boeing 777), the agent recognized the missing capability and filed a structured bug report with reproduction steps and URL parameters.

110

André Bandarra @bandarra.me · 23/09/2026

When I wrote the WebMCP guidance on building tools for Chrome for Developers, the focus was on design and testing. After launch, developers face survivorship bias: we only see data for journeys that work. If a tool was missing or an agent got stuck, developers get zero signal.

110

André Bandarra @bandarra.me · 23/09/2026

🔄 Every software dev cycle relies on a feedback loop: design, build, test, launch, and review. When users hit broken flows on a site, some file support tickets. But when AI agents hit dead ends, they fail silently or apologize in prose. What happens if we give agents a bug-reporting tool? 🧵

121

André Bandarra @bandarra.me · 22/09/2026

Microsoft Edge's #WebMCP implementation is ready for testing, and the team has new demos! Can't get enough pizza related WebMCP demos! 🍕🍕🍕

062

André Bandarra @bandarra.me · 02/09/2026

It's amazing to see agents starting to deliver value in real-world applications using WebMCP!

000

Reposted by André Bandarra

Wes Bos @wesbos.com · 25/08/2026

WebMCP landed in ChatGPT today Websites expose tools, your agent can use them, or you can just use the app like a human. Probably both! I'm calling it Clicks-n-clankers™

5252

André Bandarra @bandarra.me · 10/08/2026

Are you tunneling or building native WebMCP tools? How are you handling this in your projects? Let me know your thoughts! #WebMCP #AgenticWeb #MCP

100

André Bandarra @bandarra.me · 10/08/2026

We need better ways for WebMCP and MCP servers to coordinate, especially for complex, multi-step tasks where tools are split between both, and a standardized way for agents to discover MCP servers from a site, without just falling back on tunneling.

100

André Bandarra @bandarra.me · 10/08/2026

Plus, tunneling can cause agent confusion. When agents discover the same tools via different "Connectors," they end up with duplicate tools, causing bloat and making it hard for the agent to know which to use.

100

André Bandarra @bandarra.me · 10/08/2026

For example, a native 'add-to-cart' tool can update your shopping cart UI in real-time, or a 'set_filters' tool can adjust your site based on agent prompts. This is far beyond just tunneling.

110

André Bandarra @bandarra.me · 10/08/2026

The real power of WebMCP is deep UI integration. Because tools are part of your frontend, they can read/write directly to the DOM, cookies, and local storage.

110

André Bandarra @bandarra.me · 10/08/2026

Tunneling uses WebMCP as a conduit to expose MCP capabilities directly to the web. The request/response structures remain unchanged. It's a valid use case, but is it the best one?

120

André Bandarra @bandarra.me · 10/08/2026

Developers often use WebMCP just as a tunnel to existing MCP servers. Is this the best approach? Let's discuss. 🧵

100

Reposted by André Bandarra

Sarah Drasner @sarahedo.bsky.social · 30/07/2026

🛠️ I've made a useWebMCPTool hook for React! Chrome (myself and others) will keep this up to date with any spec changes. The hook has hygiene around unmount, result normalization, etc. Cheers! www.npmjs.com/package/use-... github.com/GoogleChrome...

npmjs.com

2175

André Bandarra @bandarra.me · 20/07/2026

What are you building, what was easy, what's blocking you from making it better. Whether it's token costs, trust on the LLM output, lack of documentation / best practices on how to build, lack of APIs on the platform, I'm interested in understand how we can help.

000

André Bandarra @bandarra.me · 20/07/2026

Want to share your experience? Drop a comment below or send me a DM, and let’s find 15 minutes to connect! 📩👇 #AI #BuildInPublic

150

André Bandarra @bandarra.me · 20/07/2026

If you’re anywhere in the trenches of the #AgenticWeb, I’d love to hear what you’re seeing and learning firsthand. 💬

100

André Bandarra @bandarra.me · 20/07/2026

I'm looking to understand: • What’s going surprisingly well? 🚀 • What are the biggest blockers holding back deep integration? 🚧 Whether it’s technical friction (APIs, docs, speed) or business hurdles (token costs, user trust, ROI)—every insight helps.

100

André Bandarra @bandarra.me · 20/07/2026

I want to learn from people in my network building AI agents into their sites. 🤖💡 Whether you’re building a custom agent, integrating a third-party tool, or creating an agent platform for others—I’d love to chat! 🧵👇

282

André Bandarra @bandarra.me · 19/07/2026

RAG? Search related stuff on your database and add to the prompt.

120

André Bandarra @bandarra.me · 06/07/2026

AI agents are changing how users interact with the web. Is your site ready? 🤖🌐 To participate in the Agentic Web you need "agent-readiness." Focus on Discovery, Understanding, and Usage to make your site accessible for AI. Read more at bandarra.me/posts/the-ag... #AgenticWeb #WebMCP #AI

bandarra.me

The Agentic Web

Unlock the future of web interaction by making your site "agent-ready" for the surging wave of AI agents. This essential guide breaks down the Agentic Web into a pragmatic framework of Discovery, Unde...

111

Reposted by André Bandarra

Paul Kinlan @paul.kinlan.me · 28/06/2026

I've built an extension for Pi that lets you communicate with a running instance of pi via discord, telegram, email and web hooks I'd love some early testers to help me with on-boarding and testing. You can reply or DM and I can send you the link.

381

Reposted by André Bandarra

Niels Leenheer @html5test.com · 25/06/2026

Really insightful workshop about WebMCP at #GoogleIOConnect - this actually might make a lot of sense to integrate in our appointment booking product. What makes it useful is that agents do not have to use the same flows as people and can answer questions like: “when is the next available time?”

152

André Bandarra @bandarra.me · 12/06/2026

In the spec there is a `requestUserInput` method the site developer can call to request permission, so more like per-inference. You can read the discussion in this issue github.com/webmachinele... This is not implemented in any browser yet, and now is a great time to give feedback on the design.

github.com

Elicitation · Issue #21 · webmachinelearning/webmcp

Gathering thoughts on supporting MCP elicitation since this would be a good way to bring the user's attention to a tab if the agent determines that their input is needed. This has also been discuss...

100

André Bandarra @bandarra.me · 12/06/2026

The site has control over the tool execution and can use that to confirm with the user before execution. More options being being discussed in the spec repo. Would love to chat more about permission scoping. Happy to book a 1:1 chat or you can file issues at github.com/webmachinele....

github.com

GitHub - webmachinelearning/webmcp: 🤖 WebMCP

🤖 WebMCP. Contribute to webmachinelearning/webmcp development by creating an account on GitHub.

110

André Bandarra @bandarra.me · 10/06/2026

Whether your are building an AI Agent that uses WebMCP tools or exposing tools on your site to AI Agents, there's security guidance to take in consideration! #WebMCP #AgenticWeb

120

André Bandarra @bandarra.me · 09/06/2026

Are you building with Angular and wondering how use WebMCP to expose tools to AI Agents? The Angular team has got you covered and published detailed docs on how to use WebMCP with the framework! Read all about it at: angular.dev/ai/webmcp #WebMCP #AgenticWeb

140

Reposted by André Bandarra

Chrome for Developers @developer.chrome.com · 29/05/2026

Testing generative AI is difficult when one input leads to thousands of unpredictable answers. We published a guide on WebMCP evals to help you test LLM touchpoints. Verify that agents understand your schemas and can successfully complete full user journeys → goo.gle/4ftQUZI #GoogleIO

171

André Bandarra @bandarra.me · 23/05/2026

Well... yeah... I mixed things, will put that down to not being a native speaker :P. Both MTTR and MTBF are shorter. Shorter MTTR is good, but shorter MTBF is bad.

010

André Bandarra @bandarra.me · 23/05/2026

Have you seen the recent discussion AI increasing "mean time to recovery" while "mean time between failures" is increasing? IMHO, increased MTBF is a signal of lower quality software, despite increased velocity to fix issues.

110

André Bandarra @bandarra.me · 22/05/2026

@tomayac.com 's I/O talk has some great ideas for incorporating AI features into a web app, using the Built-in AI APIs. I'll definitely steal some of those ideas and put them in the admin interface for my personal blog Watch the talk here: www.youtube.com/watch?v=ddBx... #WebAI #BuiltinAI

youtube.com

Build new features using built-in AI in Chrome

YouTube video by Chrome for Developers

042

André Bandarra @bandarra.me · 21/05/2026

@kulikowski.bsky.social 's I/O talk is a treasure trove to prepare your site for AI agents: 1. How agents 'see' your site, best practices to prepare, and developer tooling. 2. WebMCP deep dive with cool demos, Puppeteer & DevTools support. Watch: www.youtube.com/watch?v=HdCc... #AgenticWeb #WebMCP

youtube.com

Build your website for the agentic era

YouTube video by Chrome for Developers

182

Reposted by André Bandarra

Sarah Drasner @sarahedo.bsky.social · 20/05/2026

🛠️ Ultimately, WebMCP is about making a more secure agent pathway instead of scraping DOM and a11y trees. I made a small site to demo how it works, how agents use it, what the spec looks like, and some important links. It's interactive! You can use the extension to actuate the page:

4648

Reposted by André Bandarra

Chrome for Developers @developer.chrome.com · 19/05/2026

The Prompt API is now in stable with Chrome 148. Use it with Gemini Nano to build web AI experiences with multimodal inputs and structured output for richer experiences. Plus, you'll now have access to expanded language support → goo.gle/43hL06y

17172

Reposted by André Bandarra

Kenji Baheux @exploractical.com · 20/05/2026

🚀 Google I/O updates for Chrome built-in AI! 🎉 Prompt API ships to stable in Chrome 148! Previews on the horizon: 🛠️ Native Tool Calling preview 🔍 Embeddings API preview. Also a new 🎛️ Prompt Parameters Origin Trial. Join the Early Preview Program for heads up and feedback: goo.gle/chrome-ai-de...

042

Reposted by André Bandarra

Lea Verou, PhD @lea.verou.me · 19/05/2026

Thrilled this is out, keeping quiet has been SO HARD! 🤐 I’ve had the opportunity to contribute as a domain expert consultant over the past few months, and saw firsthand the level of thoughtfulness and care that went into it. Proud to have played a small part, and excited to see what folks build! ❤️

66010

André Bandarra @bandarra.me · 19/05/2026

WebMCP will enter the Origin Trial in Chrome 149, just a few weeks away. But you can try it today by enabling a Chrome flag! Want to now more? Check out the docs at developer.chrome.com/docs/ai/webmcp

lnkd.in

This link will take you to a page that’s not on LinkedIn

052

André Bandarra @bandarra.me · 11/05/2026

I wrote a Markdown editor on the web that has an AI Agent attached to it. The agent an read, write, edit, read other documents in the workspace, delegate to other agents and more. Check it out at bandarra.me/apps/agent-t... #AgenticWeb #AI

011

Reposted by André Bandarra

Kenji Baheux @exploractical.com · 11/05/2026

Zack's demo (from @brilliantsole.bsky.social ) got my week off to a great start! 🚀 Pair smart glasses with the Prompt API and a few other technologies, and apparently, you get a digital first mate with a voice of pure authority. 🏴‍☠️ Check the vibe: youtu.be/tcqrFwadepQ?...

youtu.be

Prompt API with TTS on Smart Glasses | Brilliant Wear SDK

YouTube video by Zack Qattan

141

André Bandarra @bandarra.me · 09/05/2026

Good thing tomorrow is Sunday. I'm going to stay up late today playing some MIDI music :D

110

André Bandarra @bandarra.me · 09/05/2026

I want to try the TSOMI theme with colour notes!

110

Reposted by André Bandarra

Chrome for Developers @developer.chrome.com · 08/05/2026

It's time to wire everything together into a continuous testing pipeline → goo.gle/4tSIcZo Today, we'll cover extended unit tests for PR merges, weighted Composite Metrics for nightly integration runs, and Pairwise Evaluation to mathematically prove which model is best when you want to swap LLMs.

3143

André Bandarra @bandarra.me · 06/05/2026

AI agents often feel bolted to sites. Moving the agent loop to the client-side enables tight integration between the agent and your site's user interface, while still allowing usage of big cloud LLMs or delegating tasks to server-side agents Read more at bandarra.me/posts/bringi... #AI #AgenticWeb

000