Server has 9 tools with complete JSON Schema input definitions and descriptions in English + Thai. Naming follows verb_noun pattern (remember, recall, read_memory, forget, list_memories, etc.). However, descriptions are inconsistent in depth and clarity. Most parameter descriptions are minimal or absent. Output schemas are not documented. Error handling guidance is missing. Tool composition is reasonable but some tools lack clear use-case differentiation.

• Document output schemas for all tools. E.g., 'remember returns {id: string, created_at: ISO8601}'. Include all fields the LLM might need for chaining.
• Add descriptions to all parameters. For 'id', specify format (8-char alphanumeric). For 'limit', specify range (1-100). For 'tag', explain filtering behavior.
• Clarify tool selection criteria. Add to recall: 'Use for substring search across title/content/tags. Use list_memories to get newest N memories without search.' Add to list_memories: 'Use to browse recent memories. Use recall for keyword search.'
• Add error guidance. E.g., 'If read_memory returns null, the id may be invalid, try recall() to find the correct id.' For forget: 'Deletion is permanent. Consider archiving instead if recovery is needed.'
• Implement confirmation pattern for forget. Return {confirmed: false, id, title, content} and require a second call with confirmed=true to actually delete.

Inferred effective spec: 2026-07-28+.

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