W3C

– DRAFT –
WAI-Adapt Teleconference

23 Sep 2025

Attendees

Present
Abhinav, janina
Regrets
Annalu, Lionel, Russell
Chair
Janina
Scribe
janina, matatk

Meeting minutes

Renaming WKD

janina: Thinking 'common' or 'discoverable'

<matatk> Naming discussion from the 9th: https://www.w3.org/2025/09/09-adapt-minutes.html#59b0

matatk: Really like discoverable

matatk: Goes with the html ethos of discoverable metadata

janina: possibly also 'standard'

matatk: maybe discvoerable common destinations'

matatk: Recall us suggesting 'waypoints'

matatk: Also 'familiar' or ' core' from Lionel previously

janina: 'WAIpoints' ??

matatk: 'signposts'

Agentic PR

Abhinav: w3c/adapt#326

janina: I really like it, I think it's really well written. My only two concerns...
… (1) we're not tying any of our remit in here - everything you say I think is true, but our remit is that it's tied to accessibility, but that's not clear. Maybe we can cross-reference some requirements in here.

Abhinav: I think it's related to accessibility in that, at the end of the day, these tools help you to do jobs.

janina: I agree, but we need to make that clearer. Then the other thing...
… (2) MCP figures in there very prominently. I think maybe that's good. But is that an open protocol, or is it in IETF?

Abhinav: It's a protocol from Anthropic

Abhinav: It's open source

janina: That's a company.

janina: that may or may not be a problem; I don't know.
… beyond that I had some minor points of writing style.
… need to credit Abhinav

matatk: With respect to MCP, there are some concerns; were discussed at TAG

matatk: There are others, but this one seems to be the favorite suddenly

matatk: W3C is considering where to focus

matatk: mcp is being considered in a WG; but it's so recent

janina: Can we re-phrase and use MCP as an example. What is the (taxonomic) service that MCP is providing?

Abhinav: from perspective of agentic systems, there's a 'brain' which is an LLM and a tool that's performing what it's told.
… simple e.g. planning a vacation
… MCP is a protocol to expose those tools and resources needed.
… MCP is being used to expose the tools, but we can always talk about tools in general, and their relation to LLMs. That arrangement will persist, but protocols may change.

janina: is it a request broker?

Abhinav: can be seen a bit like that.

janina: so we could say 'a good request broker, example MCP' so we are not closing on it, but it's an example of the function.

Abhinav: OK, will try to figure out the best common word, and use MCP as exaple.

matatk: have some specific comments -- now or off line?

matatk: Reiterating some comments, esp we shouldn't standardize just for machines

matatk: but if we tie to a11y, perhaps most of the contents moves to the main explainer, or phps to the RQTF AI-A11y pepr as an example of agentic functionality

Abhinav: agree to the existing explainer

Abhinav: wonders process; should we post and then see where we want to move things?

matatk: suggests anything that might be removed at least to an appendix so we don't lose any of this wonderful work

janina: general agreement to move to a WD and then copy, move, etc, as seems best

Possible Synergies

matatk: Describing TAG discussion related ...

https://github.com/w3ctag/meetings/blob/gh-pages/2025/09-Hong-Kong/09-16-minutes.md#mcp

janina: Two kinds of simplification: content, and interaction UI.

Minutes manually created (not a transcript), formatted by scribe.perl version 244 (Thu Feb 27 01:23:09 2025 UTC).

Diagnostics

Succeeded: s/way points/waypoints/

Succeeded: s/sign posts/signposts/

Maybe present: matatk

All speakers: Abhinav, janina, matatk

Active on IRC: Abhinav, janina, matatk