Universal persistent memory system for AI tools. Stores memories, journals, sessions, and user/soul context across multiple AI applications.
mcp-memory has 15 tools with basic descriptions and input schemas, but significant quality gaps. Most tool descriptions are 20-80 chars (below the 194-char baseline for A+ tools). Parameter descriptions are minimal or missing context. No output schemas documented. Error handling is absent, tools return generic HTTP errors without recovery guidance. Tool naming is reasonable (verb_noun pattern), but composition issues exist: memory_create, soul_update, user_update, session_log, and self_update all modify state without confirmation or dry-run support. The self_update tool (IRREVERSIBLE risk) has no safety gates. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) present. Schemas are present but sparse, many parameters lack type constraints, enums, or format specifications.
Log an entry to today's journal. Use this to record decisions, observations, and progress.
Read journal entries for a specific date.
Create or update a persistent memory. Memories persist across sessions and tools.
Delete a memory by name.
Read a specific memory by name.
Grep through all memories and journals with a regex pattern. Returns matching lines with file and line number, like grep.
List all memories with lightweight index info (no full content).
No output schemas documented. Tools return JSON but LLMs cannot plan downstream calls without knowing response structure (e.g., does memory_list return {memories: [...]} or {data: [...]}?). Violates pattern:tool and mxe:response-field-naming.
Destructive/irreversible tools lack confirmation or dry-run support. memory_delete, soul_update, user_update, and self_update can cause data loss or system changes without agent confirmation. Violates pattern:confirmation-request.
Error handling is absent. HTTP handlers return generic 'Failed to X' or 'Invalid request body' without actionable recovery guidance. LLMs cannot self-correct. Violates pattern:recovery-guide and pattern:error-classification.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 49 | 2026-07-28+ | v2 |
Search memories by keyword. Returns matching memories with content.
Find memories similar to a query using keyword overlap scoring.
Update mcp-memory to the latest version. Downloads and replaces the current binary.
Identify which AI model is being used in this session. Call this at the start of every session.
Log an action to the current session for tracking.
Run this to set up mcp-memory as your primary memory system. Returns instructions you MUST follow: add rules to your tool, migrate existing memories, and switch to using mcp-memory exclusively.
Update the SOUL.md persona/identity file.
Update the USER.md file with context about the human user.
Parameter descriptions are minimal. 'Memory name', 'Query to find similar memories for', 'Date in YYYY-MM-DD format' lack context on constraints, valid ranges, or when to use. Baseline is 72 chars; most here are 20-50 chars. Violates pattern:tool-description.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). LLMs cannot infer which tools are safe to retry, which modify state, or which are read-only without explicit hints. Violates current MCP spec (2026-07-28) tool annotation pattern.