Scaffold typed, tested, deployable MCP servers from a declarative definition.
This is a code generator framework (mcp-builder), not a production MCP server. The evaluated artifact is the example dev-utils-server it generates. Three tools present: hash_text, generate_uuid, format_json. All have basic descriptions (10-50 chars, below the 50-200 char baseline for LLM optimization). Input schemas are visible in the code generation templates but parameter descriptions are MISSING, the generated code shows only type definitions without semantic guidance. No output schema documentation. No error handling guidance. The framework itself is well-structured (TypeScript, Zod validation, proper imports), but the generated server lacks production-grade tool quality.
Pretty-print or minify a JSON string.
Generate one or more random UUID v4 strings.
Compute a cryptographic hash of a string.
Parameter descriptions missing. The 'algorithm' enum in hash_text has no description explaining what each option does or when to use sha256 vs sha512 vs md5. LLMs cannot infer security implications or performance tradeoffs.
Tool descriptions are too brief (10-50 chars vs 50-200 char baseline). 'Compute a cryptographic hash of a string' lacks context on when to use this tool vs alternatives, what formats it accepts, or what the output looks like.
No output schema documentation. Tools return content arrays but LLMs have no formal schema describing the response structure, field types, or what to expect.
generate_uuid has no input schema visible. The tool accepts optional parameters (count, format) but the code generation template does not show parameter definitions or constraints.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 49 | 2026-07-28+ | v2 |
No error handling guidance. If hash_text receives invalid input or format_json receives malformed JSON, there is no documented recovery path or actionable error message for the LLM.