edgarmcp demonstrates solid tool design with clear naming (verb_noun pattern), comprehensive parameter schemas, and good descriptions. All 5 tools have documented input schemas with types and descriptions. However, output schemas are not explicitly documented in the source, and error handling guidance is minimal. Tool descriptions average ~150 chars (within baseline 194), but lack dependency hints and recovery guidance. Parameters are well-typed but some lack format constraints (e.g., date strings, accession number format). No tool annotations (readOnlyHint, destructiveHint) despite all tools being READ_ONLY. Missing pagination documentation for list-returning tools.

• Add explicit JSON Schema output definitions to each tool. Document response fields, types, and whether results are paginated. Example: get_filings should document that it returns {filings: [{accession_number, form_type, filing_date, ...}], total_count, next_offset}.
• Add readOnlyHint=true to all 5 tools via FastMCP's tool annotation API. This signals to agents that calls are idempotent and safe to retry, improving reliability.
• Implement pagination for list-returning tools: add limit (default 20, max 100) and offset parameters to get_filings, search_filings, search_edgar. Return total_count and next_offset in responses.
• Add regex patterns and length constraints to schema for dates (pattern: ^\d{4}-\d{2}-\d{2}$) and accession numbers (pattern: ^\d{10}-\d{2}-\d{6}$). This prevents LLM hallucination of invalid formats.
• Enhance error messages with recovery hints. When company lookup fails, suggest 'Try search_edgar() with the company name or ticker.' When accession number is invalid, return 'Invalid accession format. Expected: 0001234567-89-123456.'