Skip to content

feat(mcp): report conservative runtime capabilities - #282

Merged
jacob-bd merged 1 commit into
jacob-bd:mainfrom
Sanity-Cloud:upstream/server-capabilities-20260805
Aug 5, 2026
Merged

feat(mcp): report conservative runtime capabilities#282
jacob-bd merged 1 commit into
jacob-bd:mainfrom
Sanity-Cloud:upstream/server-capabilities-20260805

Conversation

@insane66613

Copy link
Copy Markdown
Contributor

Summary

Extend server_info with conservative capability information derived from the built-in MCP tool registry and current visibility configuration.

The response adds:

  • mcp_capabilities: registered and visible built-in tool counts, group availability, hidden tools, missing tools, and ungrouped tools
  • provider_capabilities: an explicit not_probed result

Motivation

MCP clients need a stable way to determine which built-in operations this process can expose, particularly when tool groups are disabled. They should not infer provider/account capabilities from subscription labels or from the existence of unrelated tools.

Safety and compatibility

  • all existing server_info fields are retained
  • capability reporting is read-only
  • group availability is true only when every configured tool in the group is registered and visible
  • provider-side features, quotas, plans, and secure-computer access are explicitly not inferred
  • optional external/plugin tools are not claimed as built-in capabilities
  • no new dependencies

Validation

  • uv run ruff check .
  • uv run ruff format --check .
  • focused capability suite — 4 passed
  • manual server_info smoke: 43 registered, 43 visible, no ungrouped built-ins, provider status not_probed
  • full upstream-first integration stack on Windows — 1,318 passed, 39 skipped

@jacob-bd
jacob-bd merged commit 40374b5 into jacob-bd:main Aug 5, 2026
3 checks passed
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