A comprehensive setup utility for building resilient, multi-tiered memory architectures for OpenClaw agents.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install agent-memory-setup-qmd
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 agent-memory-setup-qmd 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 Agent Memory Setup?
The Agent Memory Setup skill is designed to equip Openclaw Skills agents with a robust, hierarchical memory infrastructure. It moves beyond simple chat history by implementing a 3-tier system: HOT memory for active session state, WARM memory for stable preferences and technical references, and COLD memory for long-term archival and distilled lessons. This ensures agents maintain context across sessions without suffering from amnesia during long-running tasks.
By integrating semantic search via QMD and context compaction through Lossless Claw, this skill allows agents to handle massive amounts of information efficiently. It automates the creation of directory structures, daily logs, and heartbeat monitoring files, providing a standardized environment that makes any OpenClaw agent more reliable and context-aware.
Agent Memory Setup Use Cases
- Onboarding a new AI agent into a fresh OpenClaw workspace.
- Implementing semantic search capabilities for agents managing large documentation sets.
- Preventing context window overflow in long development or research sessions.
- Standardizing memory structures across multi-agent environments for shared context.
- Setting up automated daily logging and periodic memory distillation tasks.
How Agent Memory Setup Works
- The setup script initializes a standardized directory structure including specialized folders for hot and warm memory tiers.
- Core memory files (HOT_MEMORY.md, WARM_MEMORY.md, and MEMORY.md) are generated with protective headers to prevent accidental overwrites.
- An AGENTS.md template is deployed to the workspace to instruct the agent on how to interact with its new memory tiers.
- The system attempts to install and configure QMD for semantic search and the Lossless Claw plugin for intelligent context compaction.
- Global or agent-specific configuration files are updated to enable heartbeat checks and context pruning safeguards.
Agent Memory Setup Setup
To initialize the memory system within an OpenClaw workspace, execute the following commands:
# Run the core setup script providing your agent workspace path
bash scripts/setup_memory.sh /path/to/workspace
# Install required plugins if not already present
openclaw plugins install @martian-engineering/lossless-claw
pip install qmd
# Restart the OpenClaw gateway to apply configuration changes
openclaw gateway restart
Agent Memory Setup Data Schema & Taxonomy
The skill organizes data into a tiered hierarchy to balance retrieval speed and storage depth:
| Tier | Storage Location | Data Type | Retention |
|---|---|---|---|
| HOT | memory/hot/HOT_MEMORY.md |
Active tasks, pending actions, current session state | Volatile/High-frequency |
| WARM | memory/warm/WARM_MEMORY.md |
User preferences, API references, recurring configuration | Persistent/Medium-frequency |
| COLD | MEMORY.md |
Historical milestones, architectural decisions, core lessons | Permanent/Low-frequency |
| LOGS | memory/YYYY-MM-DD.md |
Raw session event logs and granular interaction history | Daily |
Agent Memory Setup Advanced Features
- Lossless Claw Integration: Automatically compacts old conversation history into expandable summaries to stay within context limits without losing data.
- Semantic Search (QMD): Enables agents to perform vector-based searches across the entire memory directory for rapid information retrieval.
- Heartbeat Monitoring: Configurable triggers that prompt the agent to perform self-maintenance, such as reviewing daily logs or promoting items to COLD memory.
- Multi-Agent Support: Built-in protocols for shared workspaces where multiple Openclaw Skills agents read from a common memory pool.
SKILL.md
Loading
Related Openclaw Skills
Chinese De-AI Content Refiner (zh_de_ai_ify)
A specialized tool for stripping artificial AI-smell and corporate jargon from Chinese text to produce natural, human-sounding content.
Achievement-QZTC Analysis Generator
A professional tool for generating and auditing course objective achievement analysis reports from Excel data.
TickTick CLI Skill
A comprehensive command-line interface skill for managing TickTick tasks, projects, and reminders through automated workflows.
Dida CLI Task Management
A powerful command-line interface for managing Dida365 tasks and projects through automated AI agent workflows.
ExpertPack Eval
A specialized evaluation toolkit to quantify the esoteric knowledge density and performance benchmarks of your AI knowledge packs.
Ontology to ExpertPack Converter
A technical migration utility designed to transform append-only Ontology knowledge graphs into portable, high-fidelity ExpertPacks.
Featured*
