Loading ToolBench
AI-native second brain. Self-hosted, token-optimized memory for AI assistants.
mgi-mind provides 8 tools with consistent naming (verb_noun pattern: mind_memory_add, mind_memory_search, etc.) and complete input schemas with typed parameters. However, descriptions are generic and lack LLM-optimization guidance. Most tool descriptions are 30-50 chars, below the 50-200 char baseline for A+ tools. Parameter descriptions exist but are minimal (e.g., 'Search query' for mind_memory_search's query param). No output schemas documented. Error handling and recovery guidance absent. Tools follow single-responsibility principle but lack composition hints (e.g., which tools chain together). Security considerations (permission gates, audit trails) not visible in definitions.
Add a structured fact (subject-predicate-object triple)
Create a new memory library
List all memory libraries
Add a memory to the knowledge base
Browse memories with optional filtering
Recall memories organized by type (facts, memories, procedures)
Search memories by semantic similarity
Add a memory with provenance metadata (URL source)
Tool descriptions are too short (30-50 chars) and lack LLM-optimization guidance. Descriptions should explain WHAT the tool does, WHEN to use it vs similar tools, and any prerequisites. E.g., 'mind_memory_search' description 'Search memories by semantic similarity' does not explain when to use it instead of mind_memory_recall or mind_memory_browse.
Parameter descriptions are minimal and lack format/constraint guidance. E.g., 'tier' param in mind_memory_search has description 'Retrieval tier (0-3)' but does not explain what each tier means or when to use which. 'created_since' and 'created_before' in mind_memory_browse lack ISO 8601 format specification.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 44 | <=2025-11-25 | v2 |
No output schemas documented. LLMs cannot plan downstream tool calls or extract required fields without knowing what each tool returns. E.g., does mind_memory_search return memory_id, content, library, timestamp? Are results paginated?
No error handling or recovery guidance visible. Tools do not document what errors can occur, whether they are retryable, or what the LLM should do next. E.g., if mind_memory_search finds no results, should the LLM retry with a different query or call mind_memory_browse?
Tool composition hints missing. No documentation of which tools chain together or what IDs/references are passed between them. E.g., does mind_memory_add return a memory_id that mind_memory_search can filter by? Can mind_fact_add results be browsed with mind_memory_browse?