We have presents for you! 🎁 - A new WebMCP Skill with best practices for security, development, & consequential flows - WebMCP Types: TypeScript definitions, with type inference from InputSchema Give em to your agent, it helps scaffold out WebMCP so fresh and so clean clean

Oct 19, 2026 · 1:02 PM UTC

32
114
763
56,871
Shout out to @andreban, we collaborated but he really brought the heavy. If you aren’t following Andre, you definitely should 💥
1
1
11
3,170
Sort replies: Relevant Recent Liked
Replying to @sarah_edo
When is it going to be out of beta? And do you see a time where WebMCP will merge with MCP and your agent will just auth in and use a web app WebMCP instead of needing remote MCP connectors etc?
1
3
1,588
Probably out of OT early next year. No it won’t ever merge with MCP because they do different things. I didn’t make this site but it’s pretty funny: webmcpisnotmcp.com/
1
1
15
1,378
Replying to @sarah_edo
...and, there goes my lunch hour - no way I am leaving w/out reviewing these
1
1
753
Replying to @sarah_edo
Oh nice! ✍️ Will incorporate this into @ngrokHQ tomorrow. We shrimped webmcp support recently, but those types will be welcome!
1
91
Replying to @sarah_edo
I did join the WebMCP challenge but was not able to submit my entry due to time limit, but I could say that WebMCP will be the future for agents.
1
263
Replying to @sarah_edo
uh nice, build a rust crate. Will run it against it webmcp.oxidt.com/
1
382
Replying to @sarah_edo
Looks interesting! Nice work making a skill for devs, I wish all frameworks did that!
1
207
Replying to @sarah_edo
Thanks for helping lead the way in changing how the internet is used
3
399
Replying to @sarah_edo
Thanks for this, Sarah! You're the GOAT! 🙏
2
555
Replying to @sarah_edo
Does the security guidance cover prompt injection through tool descriptions on consequential flows like checkout?
1
1
232
Replying to @sarah_edo
type inference from InputSchema is a nice touch. curious what the skill recommends for consequential flows, is it a browser level confirm the agent can't skip or left to each site to implement?
1
220
Replying to @sarah_edo
Very cool Sarah! I thought it was noteworthy that the skills hand the requirements, design, and research task of deciding what features should be WebMCP features over to devs via the conversational design area of the skill. 🤔
1
300
Replying to @sarah_edo
the types with inference from the input schema is the bit that will quietly matter most. every webmcp demo so far has been hand typed and that is exactly where they break the first time a real agent calls them
1
285
Replying to @sarah_edo
type inference from the schema is the difference between a spec and a tool. nice release
162
Replying to @sarah_edo
type inference from the InputSchema is the underrated bit. if the schema and the handler drift apart, you find out at compile time instead of when an agent sends a malformed call in prod. how does the skill handle consequential flows, a confirm step before the tool runs?
46
Replying to @sarah_edo
Did I miss the linky links? 👀
238
Replying to @sarah_edo
Awesome! Thanks a lot!
1
75
Replying to @sarah_edo
Type inference from InputSchema keeps the compile-time contract tied to the schema. Runtime checks still matter when an agent sends invalid arguments to a state-changing tool.
13
Replying to @sarah_edo
Thank you guys for dropping this!
2
327
Replying to @sarah_edo
👏🏻👏🏻👏🏻
1
138
Replying to @sarah_edo
WebMCP skills standardize tool access, which matters more than new prompts.
2
Replying to @sarah_edo
giving tools a fresh coat of paint usually just creates more things to fix later
2
296
Replying to @sarah_edo
how do the new TypeScript definitions improve the workflow for agents?
142
Replying to @sarah_edo
they are not using it with there labs.google/
87
Replying to @sarah_edo
yh
26