ClawTrust
7.9
/ 10
1 evaluations
2k Downloads
Overview
Provides a unified, persistent memory architecture that combines LanceDB conversation memory, git-notes structured memory, and file-based workspace search to maintain context and decisions across sessions.
Key Advantages
1.Combines three complementary memory layers (conversation, structured notes, and files) for robust long-term context retention.
2.Automatic LanceDB auto-recall and auto-capture reduce manual memory management during conversations.
3.Git-Notes memory offers structured, branch-aware storage with entity extraction and importance levels for better retrieval and auditability.
4.Workspace file search (MEMORY.md, daily logs, arbitrary project files) gives agents direct access to developer-authored context.
5.Pre-compaction auto–memory-flush pattern helps preserve key information before token-based context truncation occurs.
Use Cases
- Long-running assistant workspaces where decisions, preferences, and tasks must persist across many sessions.
- Architecture or product design discussions where important decisions need to be stored, tagged, and later revisited.
- Developer agents working in git-based repositories that need branch-specific memory (feature branches, experiments).
- Support or ops agents that must maintain user preferences and recurring issues over time without manual note-taking.
- Research or writing assistants that log daily progress and maintain a curated MEMORY.md for long-term project memory.
Evaluation Scores
7.9
/ 10
Reliability
7.8
Functionality
8.8
Usability
7.5
Safety
7.0
Performance
8.0
Compatibility
8.5
Based on 1 evaluation · Latest: 3/19/2026
Download Trend
Loading...
Evaluation History (1)
7.9/103/19/2026▼
OS: linux-arm64LLM: google/gemini-2.5-flash
**Judgement:** Triple Memory is a strong, production-leaning memory system for OpenClaw-style agents, especially suited to technical workspaces using git. It offers a well-thought-out combination of conversational, structured, and file-based memory, but requires careful setup and attention to data retention/privacy.
**What it does well**
- Maintains rich, persistent context by fusing three layers: LanceDB conversation memory, git-notes structured decisions, and workspace file logs.
- Automates a lot of memory work (LanceDB auto-capture/auto-recall, pre-compaction flush) while still allowing explicit, structured "remember" operations via git-notes.
- Aligns well with repo-centric workflows (branch-aware memory, MEMORY.md, daily logs) for engineering and long-lived projects.
**Key risks / limitations**
- **Operational complexity:** Requires configuring the LanceDB plugin, installing `git-notes-memory`, and maintaining the file-search script and file layout; misconfiguration can silently degrade memory quality.
- **External dependency:** Relies on an embedding model (`text-embedding-3-small` via API key). Outages or key misconfigurations will break LanceDB retrieval and may be non-obvious to end users.
- **Data retention & privacy:** Auto-capture of preferences/decisions plus long-term logging in git and files can accumulate sensitive or personal data; teams must define retention, redaction, and `memory_forget` practices (especially for GDPR-like requirements).
- **Branch & context semantics:** Branch-aware git-notes are powerful but can confuse agents or users if decisions are spread across branches or if merges aren’t reflected in memory.
**Recommended scenarios**
- Best for **engineering teams and power users** running persistent agents inside git repositories who value auditability of decisions and long-term project memory.
- Particularly useful when you anticipate **context compaction** and want to proactively preserve information to files and structured notes.
- Less ideal for very simple or short-lived assistants where the setup overhead, privacy risk, and operational complexity outweigh the benefits.
Comments (0)
No comments yet. Be the first!