Hermai MCP demonstrates solid tool design with clear naming, comprehensive parameter schemas using Zod, and good descriptions. All 6 tools follow verb_noun convention (lookup_, list_, submit_, classify_, check_, fetch_). Schemas are well-typed with min/max constraints and enums where appropriate. Tool annotations (readOnlyHint, idempotentHint) are present. However, output schemas are not documented, callers cannot see what fields to expect from responses. Error handling is present but lacks recovery guidance (e.g., 'if out of credits, call upgrade_workspace'). Parameter descriptions are solid (avg ~80 chars) but some lack format hints (e.g., fetch_schema's 'params' object lacks structure guidance).

• Document output schemas for all tools. Example: 'lookup_schema returns {schemas: [{id, domain, endpoint, verified, category, description}], total_count}' so LLMs know what fields to extract.
• Add recovery guidance to fetch_schema error description: 'If 402 (out of credits), call the upgrade_workspace tool or suggest user visit upgrade_url from the error response.'
• Clarify fetch_schema 'params' parameter: 'Optional endpoint parameters as key-value pairs. Keys must match the endpoint's documented query/body parameters. Example: {limit: 10, offset: 0}.'
• Expand classify_browser_workflow description to explain use case: 'Call this first to understand whether a workflow requires browser automation or can use direct API calls. Returns one of: direct_api, hidden_endpoint, browser_only, needs_auth.'
• Add a dependency hint to lookup_schema: 'If you only have a task description, try list_public_schemas with a free-text query first, then use lookup_schema to refine by domain.'