14 tools with complete JSON schemas, descriptions, and input validation. All tools are read-only with proper annotations. Naming follows verb_noun pattern (pesquisar_, buscar_, analisar_, etc.). Descriptions are domain-specific and adequate (80-150 chars typical). Parameters have types and descriptions. Key gaps: output schemas are generic (markdown only), no pagination support despite legal domain returning large result sets, and error handling guidance is absent from descriptions.

• Define structured output schemas for each tool category. E.g., precedent search should return {tribunal, numero, data_julgamento, ementa, inteiro_teor_url, citacoes_count} as separate fields, not markdown blob. This enables agents to extract and chain data.
• Add pagination parameters (limit: 1-100, offset: 0) and document result caps in descriptions. E.g., 'Returns up to 20 precedents per page. Use offset to retrieve additional results.'
• Enhance tool descriptions with error recovery guidance. E.g., 'If no results found, try: (1) broader search terms, (2) different tribunal, (3) check spelling of legal terms.'
• Clarify tool selection logic in descriptions. Distinguish 'consultar_jurisprudenciaia' (generic, debug-capable) from 'pesquisar_jurisprudencia' (focused results) with explicit use-case guidance.
• Document timeout behavior and retry strategy for max_wait_seconds. E.g., 'If timeout occurs, the query may still be processing. Retry with same parameters after 5 seconds.'

Inferred effective spec: 2025-06-18+.

• Relies on Dynamic Client Registration (deprecated) - use Client ID Metadata Documents (CIMD)