A persistent memory framework that enables AI Agents to log operations, resume sessions, and autonomously evolve a long-term knowledge base.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install ultra-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 ultra-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 Ultra Memory?
Ultra Memory is a high-performance operation memory system designed specifically for AI Agents. It provides a robust storage layer that persists across multiple chat sessions, allowing agents to record every tool call, file modification, and technical decision. By integrating this skill into the Openclaw Skills ecosystem, developers can ensure their agents maintain a continuous understanding of long-term projects, effectively eliminating the context loss typically associated with session resets.
The system utilizes a tiered memory architecture, categorizing data into core, working, and peripheral layers to optimize context usage. It features intelligent environment detection, automatically switching between a script-based mode for local file systems and a structured internal prompt mode for web-based LLM platforms. This flexibility makes it an essential tool for maintaining consistency in complex development workflows.
Ultra Memory Use Cases
- Resuming long-term software development projects across different sessions without losing context.
- Recalling specific shell commands or technical decisions made in previous interactions.
- Automatically building a structured knowledge base of bug fixes and architectural choices.
- Developing a persistent user profile that adapts to preferred coding styles and tech stacks.
- Managing multi-agent or multi-user environments with isolated memory scopes.
How Ultra Memory Works
- The skill begins with environment probing to determine if it should run in script mode or fallback to an internal memory state.
- Upon initialization, the agent extracts the project name and resumes the most recent session state using the resume flag.
- Every significant event, such as a file write or a milestone achievement, is logged as a structured operation type.
- When a user asks about past actions, the system performs a multi-layered keyword search to retrieve relevant fragments.
- The agent triggers periodic compression cycles to summarize long conversation histories into concise milestones.
- An autonomous evolution process distills raw logs into a long-term knowledge base and updates the user profile based on observed patterns.
Ultra Memory Setup
The skill is designed to work out of the box within your Openclaw Skills environment. To manually initialize a session, use the following command:
python3 $SKILL_DIR/scripts/init.py --project your_project_name --resume
You can customize the storage location by setting the ULTRA_MEMORY_HOME environment variable, which defaults to ~/.ultra-memory/.
Ultra Memory Data Schema & Taxonomy
Ultra Memory organizes its data into structured directories and files to ensure high-speed retrieval and easy backups.
| Data Component | Format | Description |
|---|---|---|
sessions/ |
Directory | Stores operation logs and summaries indexed by unique session IDs. |
user_profile.json |
JSON | Tracks technical stack preferences, language, and observed work patterns. |
knowledge_base.jsonl |
JSONL | An append-only database of distilled technical solutions and decisions. |
summary.md |
Markdown | A human-readable overview of the current session progress and state. |
scopes/ |
Directory | Isolated memory environments for multi-user or multi-agent configurations. |
Ultra Memory Advanced Features
- Isolated memory scopes for handling multiple users or separate agents within the same environment.
- Meta-reflection logic that merges duplicate knowledge entries and identifies stale technical data.
- Time-travel recall allowing the agent to query its knowledge state as of a specific date.
- REST API server with Bearer Token authentication for remote memory synchronization.
- Local vector search integration using sentence-transformers for enhanced semantic retrieval accuracy.
SKILL.md
Loading
Related Openclaw Skills
Grants Program Marketing
A comprehensive system for designing, marketing, and managing Web3 grants programs to attract high-quality developers and community contributors.
Web3 Community Marketing
A specialized expert system for architecting, growing, and managing crypto-native communities across Discord, Telegram, and Twitter/X.
Social Listening Report
An expert system for generating comprehensive social listening reports, brand health analyses, and competitive benchmarks from digital media data.
Podcast Production Workflow
A comprehensive framework for planning, producing, distributing, and repurposing podcast content into a high-impact marketing machine.
XT Webnovel Writing Agent
A professional AI agent framework designed to generate high-quality Chinese web novels with strict anti-AI constraints and structured plot development.
ZeeLin Zhihu Auto-post
An AI-driven automation skill that researches, writes, and publishes high-quality articles directly to Zhihu.
Featured*
