Building #WebMCP tools with #TypeScript and #React?
Don’t roll your own global.d.ts files or custom useEffect hooks.
WebMCP is an experimental, fast-moving spec. Use the 2 official packages guaranteed to stay up to date instead 🧵👇
A reminder to join @kaskulikowski and I tomorrow at 5 PM CET / 8 AM PDT to chat about building effective WebMCP tools, developer feedback loops and the three homes of the agentic web.
Join us at twitch.tv/theagenticweb
My feed is full of modern motion promo videos generated with Opus 5.5 - but… HOW ABOUT THE RETRO ONES ? 😁
(Yes, we are doing a livestream with @andreban next Friday, Oct 2nd, 5pm CEST, 8am PDT)
Topics? Videos in thread 😎
What happens when a WebMCP tool needs a custom approval UI?
This post from @kaskulikowski & @liadyosef might have the exact solution to a problem I’ve been muddling through lately. 👇 (1/)
WebMCP gives agents tools from a website. MCP Apps brings UI into the agent for the human.
So what happens when a WebMCP tool returns an MCP App? 🤔
@liadyosef, co-creator of MCP Apps, and I built an experiment to find out! 1/2 🧵
Reviewing this research from @giacomozecchini was what inspired me to experiment with a #WebMCP tool that allows agent to file support tickets. Read his research and you'll see why!
This is the article that resulted from that bandarra.me/posts/webmcp-f…
Spent some time looking into how browser agents behave on websites: what they do, why they fail, and how to fix it. I know it's a bit long, but good things take time! merj.com/blog/agent-int…
