Server has 5 tools with complete input schemas and descriptions. Naming follows verb_noun convention (submit_rfq, check_rfq_status, get_quotes, compare_quotes, simulate_quote). Descriptions are substantive (100-200 chars), exceeding the 10-char minimum. However, output schemas are not documented, LLMs cannot predict response structure for planning. Parameters lack type constraints in descriptions (e.g., 'tonnage' accepts numbers but no min/max guidance). Error handling is absent, no recovery guidance or categorization. The simulate_quote tool is testing-only but lacks a clear deprecation marker. Overall: solid naming and input definitions, but missing output documentation and error patterns.

• Document output schemas for all tools. Example for submit_rfq: {"rfq_id": "string", "status": "pending", "created_at": "ISO 8601 date", "expires_at": "ISO 8601 date"}. This enables LLMs to chain calls and extract IDs.
• Add numeric constraints to parameter descriptions. E.g., 'tonnage (1.5 - 5 tons)', 'lead_time_days (1 - 30 days)', 'quantity (1 - 100 units)'. Include these in the description text since LLMs do not parse JSON Schema minima/maxima reliably.
• Implement error handling with recovery guidance. E.g., 'RFQ not found. Check the RFQ ID or call list_rfqs() to see active requests.' Categorize errors as retryable (network timeout) vs. user-fixable (invalid ID).
• Mark simulate_quote as deprecated or testing-only with explicit guidance: 'For testing only. In production, use real submit_rfq → check_rfq_status workflow. Do not mix simulated and real quotes in compare_quotes.'
• Add pagination to get_quotes and compare_quotes. Accept limit (default 20, max 100) and offset/cursor parameters. Return total_count and next_cursor in responses. Document: 'Returns up to 20 quotes per call. Use offset to fetch additional results.'

Inferred effective spec: <=2025-11-25.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry