NeveAI exposes 31 tools with minimal schema documentation. Tool definitions are inferred from builtin.py but actual parameter schemas, input types, and return structures are not visible in the provided source. Descriptions are present but generic (10-50 chars), lacking LLM-optimized guidance on WHEN to use each tool or WHAT it returns. No evidence of enum constraints, parameter validation rules, or error recovery guidance. Critical tools like execute_code and delete_memory lack confirmation patterns. The server includes dangerous operations (code execution, memory deletion) without permission gates or audit trails.

• Add explicit JSON Schema input definitions for all 31 tools. Include parameter types (string, integer, boolean, array), required fields, enums for constrained values, and min/max for numeric bounds.
• Expand tool descriptions to 100-200 characters. Answer: What does it do? When should the LLM call it instead of similar tools? What does it return? Example: 'Search user memories by keyword. Returns matching memory IDs, titles, and snippets. Use before add_memory to avoid duplicates.'
• Add parameter descriptions for all inputs. For each parameter, state its type, valid values/ranges, and purpose. Example: 'query (string, required): Search term (1-500 chars). Supports AND/OR operators.'
• Implement confirmation pattern for execute_code and delete_memory. Return a 'confirm_action' response with a summary of what will happen, requiring explicit user approval before execution.
• Document output schemas for all tools. Specify return fields, types, and pagination (limit, offset, total_count). Example: 'Returns array of {id, title, snippet, created_at}. Paginated: max 50 results per call.'

Inferred effective spec: <=2025-11-25.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry