MCP server providing episodic and semantic memory storage and recall capabilities for agents
This server has critical gaps in definition quality. All 4 tools have descriptions, but they are extremely brief (10-30 chars), falling well below the 50-200 char baseline for LLM-optimized descriptions. Parameter descriptions are missing entirely for most parameters: 'content' and 'metadata' in store_episode lack descriptions; 'limit' in recall_episodes lacks description; 'key' and 'value' in store_knowledge lack descriptions; 'key' in recall_knowledge lacks description. Input schemas are present but minimal, no type constraints, enums, ranges, or validation guidance. No output schemas are documented. Error handling is absent: the code raises ValueError for unknown tools but provides no recovery guidance. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) are present. The memory store is in-process and non-persistent, which is not documented. Tools lack context about when to use them vs. alternatives.
Recall recent episodic memories
Recall semantic knowledge
Store an episodic memory
Store semantic knowledge
Parameter descriptions are completely missing. 'content', 'metadata', 'limit', 'key', 'value' have no descriptions explaining what they control or what format they expect.
Tool descriptions are too brief (10-30 chars). Baseline is 50-200 chars. LLMs cannot determine when to select these tools vs. alternatives. No context on use cases, prerequisites, or return structure.
No output schemas documented. LLMs cannot plan downstream calls or extract required fields. store_episode returns {id}, recall_episodes returns list, store_knowledge returns 'ok', recall_knowledge returns {value}, none of this is declared in tool definitions.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 39 | 2026-07-28+ | v2 |
No error handling guidance. ValueError for unknown tools provides no recovery path. No validation of 'content' (object type but no required fields), 'limit' (no min/max bounds), 'key' (no length/format constraints), or 'value' (no type validation).
No tool annotations. store_episode and store_knowledge are destructive (WRITE risk) but lack destructiveHint. recall_episodes and recall_knowledge are read-only but lack readOnlyHint. No idempotentHint declared.