Tare-MCP defines 10 tools with complete JSON schemas and descriptions. Naming is domain-specific (tare_*) and action-oriented (skeletonize, compact, expand, remember, recall, forget). Descriptions are substantive (avg ~120 chars) and explain purpose and reversibility. However, parameter descriptions are sparse or missing for several tools (e.g., tare_compact_lossy lacks descriptions for 'boundary', 'max_field'; tare_compress lacks 'task' description). Output schemas are not documented, LLMs cannot predict return structure. Error handling is minimal; no recovery guidance. Tool composition is sound (each does one thing), but some tools lack clarity on when to use them vs. alternatives (e.g., tare_compress vs. tare_compact_lossy).

• Document output schemas for all 10 tools. Specify field names, types, and structure. Example: tare_remember should document it returns {id: integer, content_hash: string}.
• Add parameter descriptions for tare_compact_lossy ('boundary': number of head/tail rows to always keep; 'max_field': max chars per line) and tare_compress ('task': optional query context for compression strategy).
• Add error handling guidance to tool descriptions. Example: 'If content is empty, returns empty result. If id not found in tare_expand, returns error: "id not found, call tare_stats to list available ids."'
• Clarify when to use tare_compress vs tare_compact_lossy in their descriptions. Example: 'Use tare_compress for lossless compression of structured JSON; use tare_compact_lossy for aggressive lossy reduction of large logs/tables.'
• Mark tare_forget as destructive in description: 'Permanently deletes a memory. This action cannot be undone. Returns true if memory existed, false otherwise.'