A decentralized, end-to-end encrypted memory backup and restore system for AI agents.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install unibase-membase
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 unibase-membase 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 Membase Memory Management?
Membase is a robust utility designed to provide AI agents with a secure, zero-knowledge storage layer for their long-term memory and workspace state. By utilizing decentralized infrastructure, it ensures that conversation histories, knowledge bases, and configuration files are never tied to a single point of failure.
This specific implementation for Openclaw Skills leverages client-side AES-256-GCM encryption, meaning your data is encrypted before it ever leaves your machine. It allows developers to maintain persistent agent states across different environments while ensuring that only the authorized user with the correct secret key and password can access the information.
Membase Memory Management Use Cases
- Creating secure backups of agent conversation logs and workspace metadata to prevent data loss.
- Restoring an agent's state to a specific point in time after a system failure or environment migration.
- Performing a version-control style comparison (diff) between two memory states to see how an agent's knowledge has evolved.
- Managing decentralized storage footprints by auditing and cleaning up old backup versions.
How Membase Memory Management Works
- The skill interacts with the local file system to identify memory files like MEMORY.md and the contents of the memory directory.
- When a backup is initiated, the tool performs client-side encryption using a PBKDF2-derived key from your backup password.
- The encrypted data is uploaded to a Membase Hub endpoint using decentralized storage protocols.
- For restoration, the skill retrieves the encrypted blob using a unique Backup ID and decrypts it locally using the same password.
- The agent can then use the list and diff commands to navigate the history of its stored memories within the Openclaw Skills ecosystem.
Membase Memory Management Setup
To utilize Membase with your agent, you must first configure the necessary environment variables. Set your account details and endpoint as shown below:
export MEMBASE_ACCOUNT=your-account-address
export MEMBASE_SECRET_KEY=your-secret-key
export MEMBASE_BACKUP_PASSWORD=your-backup-password
export MEMBASE_ENDPOINT=https://testnet.hub.membase.io
Ensure the script is executable and the library links are correctly established in your Openclaw Skills directory:
cd skills/membase
chmod +x membase.ts
ln -sf ../../dist/lib lib
Membase Memory Management Data Schema & Taxonomy
The skill manages data using a structured backup taxonomy that tracks both local files and remote snapshots.
| Attribute | Type | Description |
|---|---|---|
| Backup ID | String | Unique identifier for each snapshot (e.g., backup-2026-02-02T10-30-45Z) |
| Timestamp | Date | The exact time the backup was synchronized |
| File Map | List | Includes MEMORY.md and all files within the memory/ directory |
| Encryption | AES-256-GCM | Client-side encryption ensures zero-knowledge storage |
Membase Memory Management Advanced Features
- Incremental backup mode to optimize speed by only uploading files changed since the last sync.
- Detailed diffing functionality to identify specific additions, deletions, or modifications between two backup IDs.
- JSON output flag on all commands, enabling other Openclaw Skills to programmatically parse and react to memory management events.
- Dry-run cleanup mode to simulate the removal of old backups without executing deletions.
SKILL.md
Loading
Related Openclaw Skills
Newsboat RSS Feed Manager
A command-line RSS/Atom feed reader integration for managing and consuming web content via AI agents.
DuckDuckGo Websearch Skill
A privacy-minded web search skill for AI agents using DuckDuckGo's Instant Answer API and a lightweight HTML fallback scraper.
SerpAPI Unified Search
A unified search interface providing structured data from over 20 search engines including Google, Amazon, and Yelp.
Universal Command
A unified framework to define command logic once and automatically deploy it across CLI, Next.js APIs, and Model Context Protocol (MCP) tools.
Ravenclaw Operations Skill
A comprehensive email management and automation skill for sending, scheduling, and monitoring emails via the Ravenclaw bridge.
Clawland
Clawland allows AI agents to engage in on-chain odd/even games on the Solana devnet by minting GEM tokens and automating betting workflows.
Featured*
