Deterministic validation of AI-generated artifacts: JSON Schema conformance, OpenAPI response conformance, SQL syntax. Metered API + MCP adapter with typed verdicts and fix hints.
Single tool 'validate' with clear naming, well-structured schema, and comprehensive parameter documentation. The tool follows verb_noun naming (validate), has a descriptive 167-character description explaining WHAT (validate artifacts against contracts), WHEN (via machinegrade API), and HOW (three validator types). Input schema properly declares types and required fields. However, output schema is undocumented, callers cannot predict what the API returns without examining the actual backend. No explicit error handling guidance or recovery hints. The server is STDIO-only, which caps protocol readiness severely but does not penalize definition quality itself.
Validate an artifact against a contract (json_schema | openapi_response | sql) via the machinegrade validate API.
Output schema not documented. Callers cannot see what fields the validate tool returns, forcing them to infer structure from error cases or upstream documentation.
No error handling guidance. Error responses from the server (mcp/server.ts lines 40-45) return raw HTTP status + text but do not categorize errors as retryable/user-fixable/fatal or suggest next steps for recovery.
Contract parameter description is generic ('Validator-specific contract'). For each validator type (json_schema, openapi_response, sql), the nested structure and required fields should be explicitly documented inline or referenced.
No pagination or result limits documented. If the validate API returns complex nested validation errors or large artifact summaries, there is no indication of size caps or chunking strategy.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | B | 74 | 2026-07-28+ | v2 |