Skip to content

sep: add progressive disclosure via groups proposal - #19

Open
helloeve wants to merge 1 commit into
modelcontextprotocol:mainfrom
helloeve:sep
Open

helloeve wants to merge 1 commit into
modelcontextprotocol:mainfrom
helloeve:sep

Conversation

@helloeve

Copy link
Copy Markdown

Adds the draft SEP for the io.modelcontextprotocol/groups extension (and links it from the README contents table).

It defines progressive disclosure of tools and other primitives: a Group is a named, lightweight bundle a server advertises once both sides negotiate the extension, with groups/list keeping listing cheap and groups/get returning a group's scoped tools, prompts, resources, and nested groups as typed, invokable definitions in a single stateless round-trip.

Scoped primitives stay out of tools/list/prompts/list/resources/list so servers can ship rich per-workflow capability without bloating every session, while scoped actions still run through ordinary tools/call (no client filesystem or shell needed).

Add the draft SEP for the io.modelcontextprotocol/groups extension
(progressive disclosure via activatable groups) and link it from the
documentation table.

@OR13 OR13 left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Isn't this challenge solved by agents and teams today?

So you don't progressively keep adding skills to the same agent, but instead, when you start hitting the limits of what separates a "specialist" from a "generalist", you make specialist teams, and make the generalist aware of how to invoke them?

Does this proposal align with giving specialist agents groups?

I guess my question is really, is there any aspect of this that interacts with agents or skills over mcp, seems like those paths could also address some of this.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants