An MCP server for persistent memory management with vector search capabilities, providing tools to save, retrieve, search, and manage contextual memories with SQLite backend and optional embedding support.
Pamie defines 8 tools with complete JSON schemas and descriptions. All tools have verb-noun naming (context_*) and documented parameters with types. However, descriptions are generic and lack LLM-optimization guidance. Parameter descriptions are minimal (10-30 chars), missing format constraints, ranges, and dependency hints. Output schemas are not documented. Error handling is absent, no recovery guidance or actionable error messages. Tools lack annotations (readOnlyHint, destructiveHint). The context_delete tool accepts a 'confirm' flag but provides no dry-run or pre-execution warning pattern.
Delete a memory entry by ID with confirmation required
Retrieve a specific memory entry by ID
List all memories with optional pagination
Retrieve recently created or updated memories
Save a new memory entry with title, body, keywords, source, metadata, importance, and pinned status
Search for memories using full-text search with optional filtering by source, metadata, creation date, and pinned status
No output schemas documented. LLMs cannot plan downstream calls or extract required fields. Tools return unstructured responses.
Parameter descriptions are too brief (10-30 chars) and lack format constraints, ranges, or dependency hints. E.g., 'importance' has no range (0-10? 1-5?), 'depth' enum values not explained.
No error handling guidance. Tools provide no recovery hints (e.g., 'ID not found, try context_search() first'). LLMs cannot self-correct on failures.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 59 | 2025-06-18+ | v2 |
Get statistics about the memory store including total count, storage size, and other metrics
Update an existing memory entry with new title, body, keywords, metadata, importance, or pinned status
context_delete lacks confirmation or dry-run pattern. Destructive operations should support pre-execution validation to prevent accidental data loss.
Tool descriptions are generic and do not explain WHEN to use each tool vs. alternatives. E.g., context_search vs. context_list distinction unclear.