Minimal instrumented MCP server for locating lost tool calls in Claude Desktop
Four tools with complete input schemas and clear descriptions. Naming is verb-based (ping, echo, sized_echo, slow_echo) and unambiguous. Descriptions are concise (20 - 80 chars) and explain purpose well. However, parameter descriptions are minimal or absent for some inputs (e.g., 'size_kb' lacks range guidance; 'delay_ms' lacks units clarity in description). No output schema documentation. Tools are trivial by design (debug/instrumentation purpose), which limits composition concerns but also means no real error handling patterns or recovery guidance. Risk classification is READ_ONLY across all tools, appropriate for a debug server.
Returns its text argument unchanged. Use a different text per chat window to tell the windows apart in the capture.
Smallest possible round trip. Returns a fixed string and the sequence number of this call.
Returns size_kb kilobytes of filler. For probing buffer and message size effects.
Returns its text argument after delay_ms milliseconds. Separates a slow handler from a lost message.
Parameter descriptions lack constraint guidance. 'size_kb' (0 - 8192) and 'delay_ms' lack explicit range/unit documentation in the description text itself; LLMs cannot reliably parse JSON Schema constraints and rely on description text.
No output schema documented. While tools return simple text content, the response structure (CallToolResult with content array) is not documented for the LLM. This is minor for debug tools but violates pattern:tool baseline.
No error handling or recovery guidance. Tools throw errors (e.g., 'Unknown tool') but do not provide actionable recovery steps. For a debug server this is acceptable, but production tools must guide LLM recovery.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 60 | 2026-07-28+ | v2 |