A sophisticated five-layer memory architecture for AI agents that enables episodic recall, semantic compression, and intelligent information forgetting.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install memcore
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 memcore 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 MemCore Enhanced: Advanced AI Memory System?
MemCore Enhanced is a high-performance memory management framework designed to provide AI agents with a structured, human-like memory system. By expanding on the original MemCore architecture, this version introduces episodic memory for event-based tracking and semantic compression to manage context window efficiency. It allows developers using Openclaw Skills to build agents that truly learn from interactions, maintain long-term context, and resolve information conflicts autonomously.
The system organizes data into five specialized layers: Context, Task, User, Knowledge, and Experience. This layered approach ensures that immediate conversation flow is maintained while long-term preferences and external knowledge remain accessible. It is an essential component for building production-grade agents that require cross-session persistence and sophisticated data retrieval.
MemCore Enhanced: Advanced AI Memory System Use Cases
- Recalling specific conclusions from previous discussions or project meetings using episodic memory.
- Reducing token usage in long-running conversations through automated semantic compression.
- Building personalized user experiences by persisting preferences and habits across multiple sessions.
- Automatically loading relevant documentation or project context when specific keywords are detected.
- Maintaining a clean and relevant memory state by filtering out low-value data using the forgetting curve model.
How MemCore Enhanced: Advanced AI Memory System Works
- The agent captures user input and evaluates it through the process_input method to identify potential memory triggers.
- New information is categorized and stored in the appropriate layer, such as User Memory for preferences or Episodic Memory for event conclusions.
- When the context window reaches capacity, the Semantic Compression engine summarizes older messages to preserve core meaning while freeing up tokens.
- During retrieval, the system uses both keyword triggers and vector-based semantic search to find the most relevant historical data.
- Periodically, the Forgetting Curve cycle calculates a health score for each memory based on age and access frequency, pruning low-importance data.
- Any conflicting information (e.g., a changed preference) is handled by the Conflict Resolver using pre-defined strategies like time or confidence priority.
MemCore Enhanced: Advanced AI Memory System Setup
To get started with MemCore Enhanced, first install the necessary dependencies:
pip install numpy
Initialize the memory core with a persistent database path to ensure your agent retains information across restarts:
from memcore_enhanced import MemCoreEnhanced
mc = MemCoreEnhanced(db_path="~/.memcore/memory.db")
MemCore Enhanced: Advanced AI Memory System Data Schema & Taxonomy
MemCore Enhanced organizes data into a structured taxonomy across persistent and volatile storage:
| Layer | Description | Retention Type |
|---|---|---|
| Context | Current session messages and sliding window | Ephemeral/Token-limited |
| Task | State machine tracking for active goals | Session-based |
| User | Persistent user profiles and preferences | SQLite Persistent |
| Knowledge | Vectorized external data for RAG | Vector Store |
| Experience | Long-term execution history for reinforcement | Persistent |
| Episodic | Specific event summaries and conclusions | SQLite Persistent |
Each memory entry includes metadata such as a Confidence Score (0.0 to 1.0), MemorySource (Explicit, Inferred, etc.), and a Priority Level.
MemCore Enhanced: Advanced AI Memory System Advanced Features
- Episodic Recall: Automatically converts past discussions into natural language summaries for the agent to review.
- Ebbinghaus Forgetting Curve: Simulates human memory decay to optimize database performance and relevance.
- Memory Triggers: Custom keyword-based hooks that automatically load specific memory clusters into the active context.
- Conflict Resolution: Intelligent strategies (Merge, Time Priority, Confidence Priority) to handle contradictory information.
- Vector Knowledge Integration: Support for semantic similarity search using customizable embedding functions.
- Confidence Tracking: A dedicated tracker that monitors the reliability of memories based on user verification or contradictions.
SKILL.md
Loading
Related Openclaw Skills
Deal Hunter Alert Claw
A high-speed information capture tool that monitors multiple platforms for low-price goods, real estate, and bidding opportunities with instant alerts.
Hello Hitch Skill
An AI-powered passenger assistant for Hellobike Hitch that manages price comparison, ride booking, and real-time trip tracking.
Jovay Interaction
A comprehensive toolset for interacting with the Jovay Layer2 EVM blockchain and Ethereum network via command-line operations.
Security Auditor (security-auditor-tk)
A comprehensive security auditing toolkit for Linux servers, web applications, and cloud environments that generates prioritized fix reports.
Jovay DApp Generation Skill
A comprehensive development toolkit for building, deploying, and debugging full-stack decentralized applications on the Jovay Layer 2 blockchain.
Humanizer-zh: AI Content De-slop & Naturalizer
A specialized editing skill that identifies and removes robotic patterns from Chinese AI-generated text to ensure it sounds natural and authentic.
Featured*
