André Bandarra retweeted
Why do we need WebMCP when a backend API already does the job?
@kaskulikowski tackles your biggest questions on session auth, browser-use, and production use cases in 90 seconds ⏱️
@kaskulikowski tackles your biggest questions on session auth, browser-use, and production use cases in 90 seconds ⏱️
3
13
133
11,907
André Bandarra retweeted
I suspect these people would be interested in this conversation @yoavweiss, @0xidanlevin, @victorhuangwq, @kaskulikowski, @andreban, @AlexNahasDev, @_jason_today, @kashhora, @naseemalnaji - I hope you can join
0
1
2
158
André Bandarra retweeted
Everyone's building something. So what are we now: knowledge workers? builders? developers? 🤔
Also... WebMCP outputSchema just dropped! 📦 😁
Join @andreban and me this Friday, Nov 6th, 5pm CET, 8am PDT - twitch.tv/theagenticweb - and let's brainstorm this!
Also... WebMCP outputSchema just dropped! 📦 😁
Join @andreban and me this Friday, Nov 6th, 5pm CET, 8am PDT - twitch.tv/theagenticweb - and let's brainstorm this!
7
2
9
549
Oh, I hadn't realized that @CodePen added #WebMCP support.
I agree with the “bring your own AI” rationale, allowing users to use the AI they already have a subscription for or is already connected to other applications.
Worth reading their blog post: blog.codepen.io/docs/pens/code…
I agree with the “bring your own AI” rationale, allowing users to use the AI they already have a subscription for or is already connected to other applications.
Worth reading their blog post: blog.codepen.io/docs/pens/code…
2
1
12
1,196
outputSchema just landed in the #WebMCP specification (github.com/webmachinelear…), letting web apps declare the JSON Schema of what a tool returns alongside inputSchema.
It closes a subtle gap when designing tools for larger apps.
It closes a subtle gap when designing tools for larger apps.
2
6
18
2,464
outputSchema separates intent from structure: description stays short and focused on when to call the tool, while outputSchema maps out the returned fields up front so the agent can pick and chain the right tools on the first try.
1
0
1
62
Testing identical 17-tool schemas with and without outputSchema raised pass rates from 5.6% to 88.9% and cut tool calls by 42%.
Post: bandarra.me/posts/webmcp-o…
Post: bandarra.me/posts/webmcp-o…
0
0
2
38
"How's AI changing your day-to-day work" is my new conversation starter.
I just asked the guy doing annual service on my heat pump and he told me how he takes photos and quick notes while doing the work and AI turns them into full reports, and saves him a lot of time.
I just asked the guy doing annual service on my heat pump and he told me how he takes photos and quick notes while doing the work and AI turns them into full reports, and saves him a lot of time.
2
0
9
235
Turns out, he also built a web app that is a dashboard to track a larger job using Claude Code. It's local only, and he hasn't deployed anywhere, and doesn't know how to do it, but wants to. No previous software engineering experience. I offered helping with that :D
0
0
1
50
André Bandarra retweeted
So bullish on WebMCP
wterm now supports WebMCP
Agents can use the terminals on your page
→ List, read, type, press keys
→ Opt-in per terminal, read-only mode
→ React, Vue, Svelte + vanilla
→ No DOM scraping or fake key events
Below: 5 fx agents launch each other across 4 browser terminals
112
30
656
53,177
André Bandarra retweeted
Want WebMCP tools on your website just like Shopify?
1) Copy the prompt: webmcp.com/?path=build
2) Give it to your coding agent (it uses Google's WebMCP Skills to build tools)
Win.
1) Copy the prompt: webmcp.com/?path=build
2) Give it to your coding agent (it uses Google's WebMCP Skills to build tools)
Win.
Webmcp is so much better then browseruse
3
13
193
20,102
André Bandarra retweeted
What if you connected a coding agent to your browser and let it catch, fix, and re-test broken JavaScript code automatically? 🤯
Join @kaskulikowski as he uses Chrome DevTools for agents to debug WebMCP tools live, via DevTools and the CLI.
Join @kaskulikowski as he uses Chrome DevTools for agents to debug WebMCP tools live, via DevTools and the CLI.
12
20
166
14,503
Vertical Tabs are really awesome for tab hoarders like me :D
If you have too many tabs in @googlechrome, I have something that may help!
It’s a feature called Vertical Tabs, which moves tabs from a horizontal strip on the top of the browser to a vertical panel on the left. This ensures that you can always read the title of the tab, even m open.
When I posted about it a few weeks ago, hundreds of people messaged me asking how to set it up, so I recorded a quick 2-minute video to show you.
Check it out and I look forward to your feedback!
1
0
7
480
André Bandarra retweeted
What a heartbreaking week it has been. Sir Patrick Stewart was a friend, a colleague, and a magnificent actor. From the stage to the bridge of the Enterprise, he brought intelligence, grace, and a deep sense of humanity to everything he did. Captain Picard stood for the best of us, and so did the man who played him. I was filled with affection and admiration for him and his work.
My heart goes out to Sunny and his family. He will be deeply missed.
My heart goes out to Sunny and his family. He will be deeply missed.
2,493
25,589
188,948
2,779,365
Coding with agents feels less like building with LEGO and more like working with clay.
Kind of like going from box modeling in Blender to sculpting in ZBrush.
You start with a rough lump and keep shaping it until it looks right.
#AICoding
Kind of like going from box modeling in Blender to sculpting in ZBrush.
You start with a rough lump and keep shaping it until it looks right.
#AICoding
5
1
10
643
I'm sure @ThiagoUnreal will appreciate the analogy between Blender and ZBrush here :D
1
0
1
76
André Bandarra retweeted
Unit tests check your code deterministically. But what happens when an LLM agent calls your WebMCP tools? 🎲
As Maud Nalpas explains, agent behavior varies. That's where evals come in! We built an experimental WebMCP Eval CLI tool to help you track pass rates and debug agent tiple runs.
Have you tried running evals on your WebMCP tools?
As Maud Nalpas explains, agent behavior varies. That's where evals come in! We built an experimental WebMCP Eval CLI tool to help you track pass rates and debug agent tiple runs.
Have you tried running evals on your WebMCP tools?
15
24
160
12,050
André Bandarra retweeted
You registered a WebMCP tool, but does the browser actually see it with the name, description, schema, and in the state you intended?
In Chrome DevTools, Application > WebMCP shows the tools your page exposes and lets you test them.
@kaskulikowski walks you through it ⤵️
In Chrome DevTools, Application > WebMCP shows the tools your page exposes and lets you test them.
@kaskulikowski walks you through it ⤵️
4
20
162
13,777