526201bbc8The WG's proposed priorities for the core-maintainer roadmap, per the Jul 27 meeting. Three tracks, in priority order: 1. **Graduate the Server Card as an official MCP extension** — gated on merging an SDK reference impl (Python [python-sdk#2951], Go [go-sdk#1024], TS [typescript-sdk#2527]), landing best practices ([#40]/[#36]), and keeping the GitHub showcase ([#34]) in sync. 2. **Describe what a server does, not just where to reach it** — the primitives / static-capability work ([#30]), framed around the problem: a card lists URLs but no operations for consumers that never connect. 3. **Authentication** — express real-world per-remote auth ([#13]); land a limited shape first, expand as a fast-follow. Kept deliberately short as a strawman for public comment. Docs-only; `format:check` passes. [python-sdk#2951]: https://github.com/modelcontextprotocol/python-sdk/pull/2951 [go-sdk#1024]: https://github.com/modelcontextprotocol/go-sdk/pull/1024 [typescript-sdk#2527]: https://github.com/modelcontextprotocol/typescript-sdk/pull/2527 [#13]: https://github.com/modelcontextprotocol/experimental-ext-server-card/issues/13 [#30]: https://github.com/modelcontextprotocol/experimental-ext-server-card/issues/30 [#34]: https://github.com/modelcontextprotocol/experimental-ext-server-card/issues/34 [#40]: https://github.com/modelcontextprotocol/experimental-ext-server-card/issues/40 [#36]: https://github.com/modelcontextprotocol/experimental-ext-server-card/pull/36 --------- Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com> Copilot-Session: 904e057e-7dd0-4dfc-9e15-571ddeb8475f Copilot-Session: 2eabcc66-5630-47ce-b193-171bbcaff1e8
2.4 KiB
Server Cards: Roadmap Priorities
The Server Card working group's proposed priorities for the core-maintainer roadmap. Maintainers are consolidating the roadmap around an upcoming spec release. These are the three themes, in priority order, we want carried into that roadmap.
1. Unlock MCP Discovery
The next step is to graduate the Server Card to an official MCP extension as Server Cards are still experimental. SEP-2127 is in review, and the Extensions Track (SEP-2133) also requires a merged SDK reference implementation. Getting to an accepted, adoptable extension is the top priority. This will unlock MCP as a part of a discoverable web of agents.
2. Describe what a server does
A card today says where to connect but not what the server can do. That
blocks consumers from using Server Cards for richer discovery and informing
end users of what to expect. server/discover and list endpoints are
frequently gated by auth, so an enriched server card would help significantly.
Primitives were deliberately removed from the card (PR #19); this track
revisits adding them back for static discovery.
3. Authentication
Cards need to express real-world auth well enough that a client can actually connect. Auth is per-remote and varies across endpoints, and reviewers have raised scenarios the current shape doesn't cover (#13, core-spec PR #2742).