A persistent knowledge substrate for Openclaw Skills that replaces flat-file memory with structured, confidence-scored beliefs and semantic search.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install valence-memory
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 valence-memory 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 Valence Memory?
Valence Memory transforms your AI agent from a simple conversation logger into a sophisticated knowledge worker by providing a high-fidelity substrate for long-term intelligence. Instead of relying on volatile logs, this skill implements a structured knowledge base where factual statements are treated as beliefs with dimensional confidence scores, domain classification, and strict provenance tracking. It is designed specifically to enhance Openclaw Skills by allowing agents to evolve their understanding of projects, users, and workflows over time.
By leveraging PostgreSQL and pgvector, Valence Memory enables semantic search and auto-recall, ensuring the agent always has the most relevant context without manual intervention. It doesn't just store data; it detects behavioral patterns, surfaces contradictions through tension tracking, and maintains a human-readable disaster recovery fallback through automated MEMORY.md synchronization.
Valence Memory Use Cases
- Managing long-term software architecture decisions across multiple development sessions.
- Capturing and reinforcing user preferences and recurring behavioral patterns for personalized automation.
- Resolving contradictory information in complex projects by surfacing and resolving logical tensions.
- Sharing federated knowledge and trusted beliefs across different Openclaw Skills instances.
How Valence Memory Works
- At the start of every agent turn, the auto-recall mechanism searches the database for beliefs relevant to the current prompt and injects them into the context.
- Throughout the conversation, the agent uses specialized tools to create new beliefs, update existing ones, or archive outdated information.
- When a session concludes, the auto-capture engine scans the dialogue to extract new insights and facts that were not explicitly saved.
- The system continuously calculates confidence scores based on corroboration and frequency, identifying patterns and potential tensions.
- All changes are synchronized to a local MEMORY.md file, ensuring a transparent and portable record of the agent's internal state.
Valence Memory Setup
To integrate Valence Memory with your Openclaw Skills, first install the server and database requirements:
# Install the Valence server
pip install ourochronos-valence
# Deploy the database using Docker
docker run -d --name valence-db -e POSTGRES_DB=valence -e POSTGRES_USER=valence -e POSTGRES_PASSWORD=valence -p 5432:5432 pgvector/pgvector:pg17
# Initialize and start the server
valence-server migrate up
valence-server
Then, install and configure the plugin:
openclaw plugins install @ourochronos/memory-valence
Update your openclaw.json to enable the memory slot:
{
"plugins": {
"slots": { "memory": "memory-valence" },
"entries": {
"memory-valence": {
"enabled": true,
"config": { "serverUrl": "http://127.0.0.1:8420", "autoRecall": true }
}
}
}
}
Valence Memory Data Schema & Taxonomy
The skill organizes data into a sophisticated relational schema optimized for semantic retrieval:
| Component | Description |
|---|---|
| Beliefs | Factual statements with vector embeddings, confidence scores, and provenance. |
| Entities | Defined concepts, people, or tools and their inter-relationships. |
| Tensions | Metadata tracking logical contradictions between different beliefs for agent resolution. |
| Patterns | Observed recurring behaviors or preferences detected across multiple sessions. |
| Sessions | Complete lifecycle records of agent-user exchanges and insight extraction logs. |
Valence Memory Advanced Features
- Federated Sharing: Share beliefs with trusted peers and query across distributed nodes using reputation tracking.
- Confidence-Score Explanation: Tools to help the agent or user understand why a specific belief has reached its current confidence level.
- Belief Supersession: Maintain a full version history of knowledge, allowing agents to track how facts change over time without losing history.
- Virtual Knowledge Base (VKB): Advanced session management and exchange recording for multi-agent environments using Openclaw Skills.
SKILL.md
Loading
Related Openclaw Skills
YBoard Operator
An automated assistant for managing and editing video storyboards and presentation plans on yboard.online.
SearXNG Web Search
Provides AI agents with private, aggregated web search capabilities via a local SearXNG instance.
Knowledge Base Search (kb-search)
A specialized search utility to quickly locate and filter documents within the official OpenClaw knowledge base.
Stock Analysis Skill
A comprehensive A-share market intelligence tool for real-time price tracking, automated sector screening, and personalized investment recommendations.
Feishu Multi-Agent Onboarding Skill
A specialized tool for rapidly configuring and validating multiple AI agents within Feishu via the OpenClaw framework.
Multi-Agent Hybrid Layered Isolation Architecture 1.0
A sophisticated multi-agent orchestration framework for Openclaw Skills that defines clear execution roles and isolation boundaries between specialized AI agents.
Featured*
