Lightweight TypeScript framework for dual-exposed endpoints (REST + MCP). Provides a framework-agnostic architecture supporting both REST and Model Context Protocol transports with shared endpoint definitions.
RivetBench exposes only a single 'echo' tool with severely inadequate documentation. The tool has a minimal description ('Echo') that lacks context about when to use it, what it returns, or any meaningful guidance. The input schema is present but the output schema is completely undocumented. This is a benchmark/demo server, not a production tool suite, and the single tool does not meet baseline quality standards for agent-integrated systems. No error handling guidance, no parameter validation constraints, and no composition patterns are evident.
Echo
Tool description is trivial ('Echo'), under 20 characters, provides zero context for LLM selection or use case.
Output schema is completely undocumented. LLMs cannot plan downstream operations or validate response structure.
Tool name 'echo' is generic and does not follow verb_noun convention. 'echo_message' or 'reflect_input' would be more descriptive.
No parameter description for 'message' beyond the type annotation. LLMs need guidance on expected content, length constraints, format, and purpose.
No error handling guidance. What happens if message is empty, oversized, or contains invalid characters? No recovery path documented.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 37 | <=2025-11-25 | v2 |
| 2026-03-09 | F | 41 | - | v1 |