Loading ToolBench
A cognitive memory MCP server implementing attention-gated storage and activation-weighted retrieval
Cognitive-memory has well-structured tool definitions with consistent naming (verb_noun pattern), comprehensive parameter schemas, and detailed descriptions. All 9 tools have descriptions (194 - 250 chars, within baseline 34 - 392 range). Parameters include types and descriptions. However, output schemas are not documented in the source code, LLMs cannot predict return structures. Error handling is minimal; tools lack recovery guidance. Security considerations (permission gates, audit trails) are absent from tool definitions. The 'ping' tool is a deprecated MCP pattern (removed in current spec). Overall solid foundation but missing production-grade error handling and output documentation.
Create a commitment (decision, promise, or plan) with optional deadline and concern linkage.
Write a new experience through the attention gate. The primary write path. Triggers Tier 0 synchronous processing.
Activate a concern to bias retrieval and attention scoring toward relevant memories.
Deprecate or archive a memory, removing it from active retrieval.
Get the current working context: active concerns, recent memories, and system state.
Health check tool to verify the server is running. Returns database connection status and basic system info.
Retrieve memories using semantic similarity and activation-weighted ranking.
Output schemas not documented. LLMs cannot predict return structures for any tool. No documentation of fields, types, or pagination for list results.
No error handling guidance. Tools lack recovery hints (e.g., 'If memory not found, try recall() first'). No categorization of errors as retryable vs. fatal.
Tool naming ambiguity: 'experience_tool' uses underscore+suffix pattern inconsistent with verb_noun convention. Should be 'write_experience' or 'store_experience' for clarity.
Deprecated 'ping' tool included. MCP spec (2026-07-28) removed ping; health checks should use server capabilities or dedicated health endpoints.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 49 | <=2025-11-25 | v2 |
Trigger consolidation and reflection on stored memories to extract facts, resolve contradictions, and form higher-order insights.
Explicitly create or strengthen an association between two memories.
No permission gates or security declarations. Tools like 'forget' (destructive) and 'commit' (state-modifying) lack permission checks or scope declarations (e.g., 'write:memory', 'delete:memory').