Loading ToolBench

Reading the current corpus…

hs-code-classifier-mcp-server MCP server · A · ToolBench

hs-code-classifier-mcp-server

github.com/OjasKord/hs-code-classifier-mcp-server

HS code classifier for AI agents. Classifies products to official 6-digit tariff codes before customs declarations or duty calculations. VERIFIED verdict in one call.

2025-06-18+STREAMABLE_HTTP2 tools

83/100

Definition Quality85

Protocol Readiness78

SupportabilityNot measured-

Spec posture

Inferred effective spec 2025-06-18+·current baseline 2026-07-28

Static source inference · medium confidence · evidence: structured output, Streamable HTTP

No deprecated protocol patterns detected

Summary

Two well-named tools with comprehensive input schemas and strong descriptions (150-200 chars each). Both tools declare enums for response_format and country parameters, include minLength/maxLength constraints, and provide clear guidance on input format. Output schemas are documented in descriptions. However, output structure is not formally declared in JSON Schema, and error handling guidance is minimal. Tool composition is sound (classify vs validate are distinct concerns). No security issues detected, no credentials in parameters.

Tools (2)

hs_classify_productread onlyauthsource verified85/100

Classifies a product description to official 6-digit HS tariff codes using AI-assisted matching against government tariff schedules. Returns CLASSIFIED, AMBIGUOUS, or NOT_FOUND verdict with confidence level and reasoning.

hs_validate_coderead onlyauthsource verified85/100

Validates a supplier-provided HS code against a product description using AI-powered mismatch detection. Returns VALID, INVALID, MISMATCH, or OUTDATED verdict with risk level and optional correction suggestions.

Issues (5)

HIGHtool ↗

Output schemas not formally declared in JSON Schema format. Descriptions mention return fields (hs_code, confidence, verdict, reasoning) but no structured output schema is visible in tool registration.

hs_classify_producths_validate_code

HIGHrecovery-guide ↗

Error handling lacks recovery guidance. No indication of what the LLM should do if classification fails (e.g., 'Try a more specific product description' or 'Verify country code is valid').

hs_classify_producths_validate_code

MEDIUMtool-annotations ↗

No tool annotations (readOnlyHint, idempotentHint) visible in source. Both tools are read-only and idempotent but this is not formally declared to the MCP client.

hs_classify_producths_validate_code

MEDIUMpaginated-result ↗

No pagination or result limits documented. If classification returns multiple candidate codes, no guidance on how many are returned or how to paginate.

Recommendations

Add formal JSON Schema output declarations to both tools. Example: {"type": "object", "properties": {"hs_code": {"type": "string"}, "confidence": {"type": "number", "minimum": 0, "maximum": 1}, "verdict": {"type": "string", "enum": ["CLASSIFIED", "AMBIGUOUS", "NOT_FOUND"]}, "reasoning": {"type": "string"}}}
Enhance error descriptions with recovery guidance. E.g., 'If AMBIGUOUS, try adding material or intended use to the product description. If NOT_FOUND, verify the product is a physical good (not a service).'
Add tool annotations via MCP SDK: mark both tools with readOnlyHint=true and idempotentHint=true to signal to clients that these are safe to call repeatedly.
Document result limits and pagination in hs_classify_product description. E.g., 'Returns up to 5 candidate codes ranked by confidence. Use response_format=json for machine parsing.'
Standardize country parameter schema to use minLength/maxLength instead of length for consistency with other string parameters.

Score history

Overall score trend

First recorded score · v2 rubric

83/100

Scored	Grade	Overall	Spec posture	Rubric
2026-09-23	A	83	2025-06-18+	v2