An MCP server that checks client capabilities
This server defines only 2 tools with minimal parameter requirements and no complex schemas. Both tools have descriptions that are present but lack depth regarding WHEN to use them vs alternatives, prerequisites, and actionable next steps. Parameter descriptions are largely missing (the 'format' parameter in get_client_capabilities has a description, but no other params exist to evaluate). The output schema is completely undocumented, there is no specification of what fields the response objects contain, making it difficult for LLMs to plan downstream actions. Tool naming is verb-forward ('get_*'), which is good, but both tools are READ_ONLY introspection tools with no composition opportunity or chaining pattern. No error handling is visible beyond a basic 'Unknown tool' catch-all. The server is functional but falls short of production-grade tool quality standards.
Returns a markdown table showing which MCP capabilities the connected client supports (roots, sampling, elicitation, experimental)
Returns the MCP protocol version being used by the client
Output schema completely undocumented. get_protocol_version returns JSON with clientProtocolVersion and latestSDKProtocolVersion fields, but the tool description does not specify these. get_client_capabilities returns markdown or JSON depending on format param, but no return type schema is provided.
Tool descriptions lack WHEN-to-use guidance. 'Returns the MCP protocol version being used by the client' tells WHAT but not WHEN. When should an agent call this vs get_client_capabilities? What is the user trying to accomplish? Missing context forces LLM to guess.
No input parameter descriptions beyond 'format'. The format enum ['markdown', 'json'] is described, but there is no guidance on when markdown is preferred vs json, or what the structural differences are in the output.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 46 | 2025-06-18+ | v2 |
| 2026-03-09 | F | 47 | - | v1 |
No error handling guidance. The only visible error path is 'Unknown tool', which is generic and not actionable to an LLM. What if the client has not yet initialized? What if capabilities are null? No recovery steps provided.