A persistent cognitive memory runtime for OpenClaw that provides long-term episodic and semantic storage through a FastAPI engine and Node adapter.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install smart-memory-automaton
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 smart-memory-automaton 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 Smart Memory v2?
Smart Memory v2 is a sophisticated long-term memory solution designed to transform how AI agents interact with historical data. Unlike legacy vector stores, this runtime functions as a full cognitive memory system for Openclaw Skills, enabling agents to maintain structured episodic, semantic, belief, and goal-oriented data. It bridges the gap between ephemeral chat sessions and persistent intelligence by providing a local API that handles entity-aware retrieval, reranking, and background consolidation.
By integrating this skill, developers can grant their agents a 'working memory' that survives restarts and session resets. It leverages a FastAPI backend for heavy lifting—such as embedding generation and reflection—while providing a lightweight Node adapter for seamless integration into existing Openclaw Skills workflows. The system is specifically optimized for local execution, ensuring data privacy and low-latency access to cognitive insights.
Smart Memory v2 Use Cases
- Persistent storage of user preferences, project decisions, and long-term goals across multiple chat sessions.
- Automated session summarization to capture 'session arcs' and maintain continuity in complex development tasks.
- Background reflection and conflict resolution to keep the agent's knowledge base consistent and updated.
- Context injection for complex agents that need to surface pending insights naturally during conversation.
How Smart Memory v2 Works
- The Openclaw agent initiates a query or commit through the provided tool interfaces like memory_search or memory_commit.
- A mandatory health check is performed against the local FastAPI server to ensure the cognitive engine is responsive.
- For memory commits, the system serializes the data to protect CPU embedding throughput, queuing requests if the server is temporarily unreachable.
- The background orchestrator periodically runs consolidation tasks, such as reflection and memory decay, to refine stored information.
- During response generation, active context (including projects and top-of-mind questions) is passively injected into the model's prompt.
- At the end of a session, the system extracts conversation turns and generates a summary arc to be stored as a long-term episodic memory.
Smart Memory v2 Setup
Smart Memory v2 requires a CPU-only environment to ensure compatibility across Docker, WSL, and standard laptops. Follow these steps to initialize the Openclaw Skills component:
# Navigate to the skill directory
cd smart-memory
# Set up the Python environment
python3 -m venv .venv
source .venv/bin/activate
# Install CPU-only PyTorch first (Required)
pip install torch --index-url https://download.pytorch.org/whl/cpu
# Install cognitive engine and Node dependencies
pip install -r requirements-cognitive.txt
npm install
Smart Memory v2 Data Schema & Taxonomy
The skill organizes memory into a multi-tiered taxonomy to ensure high-relevance retrieval within Openclaw Skills:
| Memory Type | Description | Key Metadata |
|---|---|---|
| Episodic | Specific events and session arcs | session_id, timestamp, tags |
| Semantic | General facts and knowledge | entity_type, importance |
| Belief | Agent or user-defined truths | confidence_score, source |
| Goal | Active objectives and tasks | status, priority, deadline |
All pending insights and retry-queued items are maintained in local JSON buffers (e.g., .memory_retry_queue.json) to guarantee data persistence during server downtime.
Smart Memory v2 Advanced Features
- Entity-aware reranking for superior retrieval accuracy compared to standard vector search.
- Automatic retry queue that flushes upon successful heartbeat or tool calls.
- Session Arc Lifecycle Hooks that automatically summarize and commit conversation milestones every 20 turns.
- Passive Context Injection that adds a structured [ACTIVE CONTEXT] block to agent prompts without manual intervention.
- Observability endpoints for real-time monitoring of pending insights and memory health.
SKILL.md
Loading
Related Openclaw Skills
Remote Terminal
A powerful remote management skill for executing commands on Linux servers via SSH, Telnet, and web terminals with built-in security auditing.
Bilibili Video Downloader
A high-performance Bilibili media toolkit for searching, downloading, and extracting metadata from the Bilibili video platform.
GitHub Push - Secure Auto-Push Tool
A specialized automation tool for secure GitHub pushing with built-in SSH detection, remote configuration, and anti-ban mechanisms.
Autonomous Agent Toolkit Automaton
A comprehensive framework for building, configuring, and orchestrating autonomous AI agents that think, act, and evolve independently.
Nine: The Cultural Significance Skill
A specialized cultural knowledge engine that explains the symbolic, mathematical, and historical significance of the number nine across Eastern and Western traditions.
Google Search WSL
A specialized tool for executing automated Google searches through a headless or windowed Chrome instance directly within a Windows Subsystem for Linux (WSL) environment.
Featured*
