A TRTC (Tencent Real-Time Communication) AI conversation application that enables real-time voice interactions with AI agents. Includes a main TRTC wrapper server and an MCP HTTP server for tool integration.
This MCP server has two tools (echo, calculate) with minimal implementation quality. Both tools have descriptions and basic input schemas, but the descriptions are extremely generic (8-15 chars), lacking any depth about when/why to use them or what they return. Parameter descriptions are sparse (single words like 'The message to echo'). No output schemas documented. No error handling guidance. No indication of idempotency, side effects, or retry safety. The tools themselves are trivial (no real business value), but even so, they fail baseline patterns for LLM-optimized descriptions and schema completeness.
Perform basic arithmetic calculations
Echo a message back to the caller
Tool descriptions are critically short (8-15 chars). LLMs cannot determine when/why to call these tools. Baseline production descriptions are 34-392 chars; these fall far below minimum.
No output schema documented for either tool. LLMs have no visibility into return structure, making downstream chaining impossible and forcing guesswork.
Parameter descriptions are trivial ('The message to echo', 'First operand'). No context about format, constraints, or why the parameter matters.
No error handling or recovery guidance. Tools return success/failure with no actionable advice for the LLM.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 40 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 38 | - | v1 |
'calculate' tool accepts 'operation' as free-form string enum but no guidance on what happens on division by zero or invalid enum values.