Three tools with complete input schemas and clear descriptions. Naming follows verb_noun pattern (get_*, find_*). Descriptions are adequate (100-200 chars) and explain purpose. However, output schemas are undocumented, LLMs cannot predict response structure. Parameter descriptions lack constraints (e.g., date format, limit bounds). Error handling is minimal (generic error objects). No tool annotations (readOnlyHint, idempotentHint). Missing guidance on when to use each tool vs. alternatives.

• Document output schemas for all three tools. Example for get_taxon: {type: 'object', properties: {scientific_name: {type: 'string'}, taxon_id: {type: 'number'}, rank: {type: 'string'}, ...}, required: [...]}. This enables LLMs to plan downstream calls and extract fields reliably.
• Add JSON Schema constraints to find_occurrences: limit as {type: 'number', minimum: 1, maximum: 100, default: 20}; startdate/enddate as {type: 'string', format: 'date'}. Constraints are machine-parseable and prevent invalid LLM inputs.
• Enhance error handling with recovery guidance. Example: if taxon not found, return {error: 'Taxon not found', suggestion: 'Try a broader genus name or call find_occurrences() to search by occurrence records'}. Classify errors as retryable (network timeout) vs. user-fixable (invalid name).
• Add tool annotations to schema. Mark all three tools with readOnlyHint: true and idempotentHint: true in the tool definition. This signals to LLMs that these are safe to call multiple times without side effects.
• Implement pagination in find_occurrences. Add optional 'offset' parameter (default 0) and return {total, count, offset, occurrences}. For large datasets, enable multi-call iteration without re-fetching.