Three tools with complete input schemas and descriptions, but quality varies significantly. add_memory has a detailed, LLM-optimized description (180+ chars) with 9 well-typed parameters and clear semantics. search_memory_facts and search_nodes have shorter descriptions (60-80 chars) and fewer parameters (2 each). No output schemas documented. Error handling is absent, no guidance on retryability, user-fixable errors, or recovery paths. Parameter descriptions are present but generic; no enums, ranges, or format constraints. Missing idempotency guarantees and pagination support for list-like tools.

• Document output schemas for all three tools. Specify the structure of facts returned by search_memory_facts (e.g., {fact_id, text, relevance_score, source}) and nodes returned by search_nodes (e.g., {node_id, name, type, description}).
• Add error handling guidance to search_* tools. E.g., 'Returns empty list if no matches found. If query is malformed, returns error with suggested corrections.'
• Implement pagination for search_memory_facts and search_nodes. Add offset/limit or cursor parameters, document max_facts/max_nodes enforcement, and return total_count in responses.
• Enhance parameter descriptions with constraints. E.g., 'max_facts: integer, 1 - 100 (default 10). Limits results to prevent context window exhaustion.' Add format hints for query strings.
• Clarify add_memory's idempotency and retry semantics. State whether passing the same uuid twice creates a duplicate or updates the existing episode. Document which parameters are immutable vs. updatable.