Bikeshed naming for the overall API

[khushalsagar]

The MS explainer called it WebModelContext while the Google one calls it ScriptTools. Filing this issue to converge on a name for the overall API.

Since the goal of the API is to align closely with MCP, "In a web browser, webpages implementing Web Model Context can easily be treated as Model Context Protocol (MCP) servers", WebModelContext sounds reasonable but its a mouthful. How do y'all feel about WebMCP?

It's too soon to bikeshed on the detailed API but wanted to nail down the overall name now so all documentation can consistently refer to this proposal.

@bwalderman @bokand @sushraja-msft

[bwalderman]

Our thinking was to name it similar to MCP so that developers who are already familiar with MCP would recognize it. My one hangup with WebMCP is the "P", since what we're building is a DOM API and not a wire protocol. That's how we arrived at just Web Model Context instead of Web Model Context Protocol. I do agree it can be a mouthful though. What do you think about just WebMC for short?

[khushalsagar]

I get what you mean about the "P" there being awkward. I still think WebMCP is better because it'll make it easier for devs to understand the connection with MCP right away. It's similar to WebGL in that aspect.

[bwalderman]

Good point about WebGL. WebMCP doesn't seem so awkward when looking at it that way. Would this mean the full name of the API is "Web Model Context Protocol"?

One other thought; we could do Web Model Context "Providers" instead of "Protocol". We get to keep using MCP as the acronym, but for folks looking more closely, it conveys that WebMCP is not a protocol and that it only does the "server" side of MCP (not the client side).

[anssiko]

It looks like we've converging on "webmcp" as the repo name.

A few considerations that support this rename, some also noted in other comments:

• By convention, the repo name will be reflected in the so-called spec shortname URL, and if/when this spec advances to the W3C's Rec Track we'd get a more concise and marketable URL: w3.org/TR/webmcp
• Once we nail down the repo name we can still fine-tune the full spec name/title (e.g. use "Providers" instead); this is not hard-coded at this stage
• MCP has name recognition among the target audience and prefixing things with "Web" is a known naming pattern

Unless I hear concerns, I'm going to rename this repo accordingly tomorrow. Rename is still cheap to do at this stage and redirects are automatic.

[MiguelsPizza]

Hey Folks, I appreciate everyone putting together this proposal. In regards to renaming to WebMCP, some attribution to the open source project that does the exact thing as the Web Model Context proposal and is already named WebMCP would be great. Particularly because I've been informed that the original hacker news post was an inspiration for this proposal .

Excited to contribute here regardless.

[khushalsagar]

we could do Web Model Context "Providers" instead of "Protocol".

Not opposed but I lean towards Protocol, it'll be less confusing. If I see "Provider" my first instinct would be: "Is it actually about MCP or they just happen to end up with the same 3 letter acronym". :)

[anssiko]

@MiguelsPizza, in a chair's capacity I'd like invite you to join the W3C Community Group so we can collaborate and develop the proposal together. The group would greatly benefit from your early implementation experience. Your web extension based implementation is a great way to quickly prototype and test ideas without the constraints of the native browser implementation. We've taken this same approach with the Built-in AI APIs also developed in this group with encouraging results: AiBrow is a web extension based implementation (with extra features), and an alternative to the native browser implementations. Certain extra features first prototyped in an extension landed into the spec and native implementations later after proving their worth with early adopters.

@bwalderman @khushalsagar, please consider adding an acknowledgments section to the explainer with a reference to @MiguelsPizza's project. It sounds like we can make the same WebMCP name work with appropriate attribution and cross-referencing to clear any confusion. We want to unite the communities. That's how we make great things happen. Exciting!

[bwalderman]

Hey Folks, I appreciate everyone putting together this proposal. In regards to renaming to WebMCP, some attribution to the open source project that does the exact thing as the Web Model Context proposal and is already named WebMCP would be great. Particularly because I've been informed that the original hacker news post was an inspiration for this proposal .

Excited to contribute here regardless.

I'm glad we're able to land in one place and start collaborating on this. The MCP-B post on hacker news came to the Edge team's attention after we began drafting the Web Model Context proposal internally but before we had a chance to publish it on MSEdgeExplainers. I wasn't aware until catching up on MicrosoftEdge/MSEdgeExplainers#1118 just recently that WebMCP was already the name of MCP-B's underlying protocol. I agree the WebMCP project merits acknowledgment in this proposal.

[anssiko]

Unless I hear concerns, I'm going to rename this repo accordingly tomorrow. Rename is still cheap to do at this stage and redirects are automatic.

Thanks for your support folks. This repo has been renamed to:

https://github.com/webmachinelearning/webmcp

PR #3 will update the explainer accordingly and will close this issue.

Looking forward to collaborating with you all on this project!

[khushalsagar]

we can close this issue now. please reopen if needed.