Local automation
Memory
Persistent local knowledge graph for assistant context across sessions.
Unranked score draft
Evidence history
Review snapshots show why the current score, confidence, and risk labels changed over time.
This review depth is not eligible for MCP Rank leaderboards until a Deep Review is complete.
| Date | Change | Score | Confidence | Risk |
|---|---|---|---|---|
| May 12, 2026 | Initial MCP Rank review captured from the curated seed dataset. Persistence can improve continuity. seed snapshot · reviewed | 77 | high | medium |
Evidence notes
Best-fit use cases
Best for long-running local assistants that need stable non-sensitive project preferences.
Useful for preserving repository conventions, terminology, and workflow notes across sessions.
Poor fit for organizations that cannot give users inspection and deletion controls over stored memory.
Risk analysis
Persistent memory can quietly accumulate sensitive context if the assistant writes secrets, customer names, or personal data.
Risk is medium even with local storage because stale or incorrect facts can influence future agent behavior.
The strongest mitigation is user-visible memory inspection plus explicit rules for what may never be stored.