Three well-named tools with complete input schemas and solid descriptions. All tools follow verb_noun naming (find, get_entry, list_database). Descriptions are 150 - 250 chars, within the 10 - 1024 baseline. Parameters have types and descriptions. However, output schemas are not formally documented, only inferred from code. Error handling returns structured objects but lacks recovery guidance. No tool annotations (readOnlyHint, idempotentHint). Missing parameter constraints (e.g., database enum, query length limits). No pagination support for list_database despite potentially large result sets.

• Add enum constraint to 'database' parameter in find() and list_database(). Replace free-form string with explicit enum: ['compound', 'drug', 'disease', 'pathway', 'genes', 'enzyme', 'glycan', 'module', 'ko'] for find(); document valid databases for list_database().
• Document output schemas in tool descriptions. E.g., 'Returns {database, query, count, truncated, results: [{id, description}]}' for find(). This enables LLMs to plan chained calls.
• Add readOnlyHint and idempotentHint annotations to all three tools in the schema. Mark all as read-only since they only call KEGG REST GET endpoints.
• Enhance error messages with recovery guidance. E.g., 'Invalid database. Valid options: compound, drug, disease, pathway, genes, enzyme, glycan, module, ko. Try find() with a valid database.' for list_database().
• Add pagination support to list_database(). Introduce optional 'offset' (default 0) and 'limit' (default 100, max 1000) parameters. Return total count and next_offset so agents can fetch all entries iteratively.