A comprehensive memory architecture for AI agents featuring tiered storage, semantic vector search, and automated memory consolidation.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install enhanced-memory-v2
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install enhanced-memory-v2 using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Memory System?
The Memory System is a sophisticated persistence layer designed for AI agents within the Openclaw Skills ecosystem. Unlike basic logging, this system utilizes a structured filesystem-based approach using Markdown files, ensuring zero database dependencies while providing powerful semantic retrieval via Ollama vector search. It allows agents to maintain long-term context across sessions by organizing information into specialized categories like User preferences and Project milestones.
At its core, the system prioritizes context management through a smart Memory Flush mechanism that triggers when token thresholds are reached. By integrating these Openclaw Skills, developers can build agents that not only remember facts but learn from human feedback over time, creating a more personalized and efficient automated workflow.
Memory System Use Cases
- Storing persistent user personas, coding styles, and technical preferences to ensure consistent AI behavior.
- Recording bi-directional feedback to help agents replicate successful patterns and avoid previous errors.
- Managing complex project progression, decision logs, and team-specific documentation.
- Indexing external reference pointers and system documentation for rapid semantic retrieval.
How Memory System Works
- Contextual Capture: The agent identifies relevant information during a session and classifies it into one of four categories: User, Feedback, Project, or Reference.
- Vector Embedding: New memories are processed using semantic embeddings (e.g., via Ollama), allowing for high-accuracy retrieval beyond simple keyword matching.
- Dynamic Loading: Upon starting a new session, the system automatically loads relevant memory fragments based on the current context.
- Persistence & Flush: When the active context window approaches a safety threshold, the system flushes memory to disk to prevent data loss or context overflow.
- AutoDream Consolidation: During idle periods, the system runs an automated maintenance task to merge duplicates, remove obsolete data, and re-index the memory files.
Memory System Setup
To install this capability from the library of Openclaw Skills, execute the following command in your terminal:
clawhub install memory-system
Ensure your configuration file (~/.openclaw/config.json) includes the following block to enable vector search and AutoDream features:
{
"skills": {
"memory-system": {
"memoryDir": "~/.openclaw/workspace/memory",
"vectorEnabled": true,
"embeddingModel": "nomic-embed-text",
"autoDream": {
"enabled": true,
"minHours": 24
}
}
}
}
Memory System Data Schema & Taxonomy
The system organizes data in a hierarchical directory structure under the designated memory directory:
| Directory | Data Type | Purpose |
|---|---|---|
user/ |
Markdown | Stores user profiles, roles, and long-term preferences. |
feedback/ |
Markdown | Contains positive/ (successes) and negative/ (corrections) logs. |
project/ |
Markdown | Tracks project-specific progress, goals, and technical decisions. |
reference/ |
Markdown | Stores pointers to external resources and system documentation. |
MEMORY.md |
Index | The main lookup index for all stored memory fragments. |
Memory System Advanced Features
- AutoDream Maintenance: An automated background process that merges redundant memories and prunes outdated information after 24 hours or 3 sessions.
- Bi-Directional Feedback: Specifically tracks both what to do (positive) and what to avoid (negative) to prevent AI over-cautiousness.
- Semantic Search API: High-level tools like
memory_searchallow agents to query their own history using natural language. - Manual Memory Flushing: Force persistence during critical operations using the
memory_flushtool.
SKILL.md
Loading
Related Openclaw Skills
Spring Fountain Revival
A robust memory survival and triple-layer backup system designed to ensure AI agent interaction data never disappears.
Diffbot Fetch
A high-performance extraction tool that converts web articles into clean, readable Markdown using the Diffbot Article API.
Whisky Encyclopedia Guide
A comprehensive AI-driven encyclopedia for exploring whisky regions, distilleries, tasting notes, and market trends.
Agent Commerce
Agent Commerce is a foundational protocol layer that enables AI agents to communicate with merchant systems for product search, price negotiation, and order fulfillment.
Memory System
A sophisticated file-based memory architecture for AI agents that combines semantic search with categorized persistent storage for long-term context retention.
Work Report Summary
A structured tool for persisting work logs in a local SQLite database to generate precise daily and weekly progress summaries.
Featured*
