Hivelore MCP has 33 tools with consistent naming (verb_noun pattern) and basic parameter schemas. However, descriptions are uniformly terse (10-40 chars), far below the 50-200 char baseline for LLM-optimized tools. Parameter descriptions exist but are minimal. Output schemas are not documented. Error handling and recovery guidance are absent. No tool annotations (readOnlyHint/destructiveHint). The server prioritizes breadth over depth, many tools exist but lack the detail needed for confident LLM selection and execution.

• Expand tool descriptions to 50-200 chars. Example: 'mem_save' → 'Save a new memory entry with optional tags and file anchors. Use this to record team policies, patterns, or decisions. Returns memory_id for later retrieval or updates.'
• Document output schemas for all tools. Example: 'mem_search returns {results: [{id, body, tags, anchors, created_at}], total_count, has_more}'. Include field types and meanings.
• Add tool annotations: mark mem_delete, mem_reject, mem_approve, scaffold_test, ingest_findings with destructiveHint=true; mark all mem_* reads with readOnlyHint=true; mark idempotent tools (mem_get, code_search) with idempotentHint=true.
• Enrich parameter descriptions with constraints and examples. Example: 'tags: array of strings (1-50 chars each, lowercase alphanumeric + underscore). E.g., ["policy", "security", "team_decision"]'. Avoid bare example values; use enums or patterns.
• Add error handling guidance to descriptions. Example: 'mem_get: Returns error if memory_id not found. Suggest mem_search() to locate the memory by content or tags. Retryable: no.'