A decentralized storage and memory chain skill for AI agents that ensures permanent history through linked-list CIDs on the Autonomys Network.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install auto-drive
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 auto-drive 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 auto-drive?
The auto-drive skill is a specialized extension for Openclaw Skills designed to provide AI agents with permanent, immutable memory. By leveraging the Autonomys Network, it allows agents to store their identity, decisions, and context as a memory chain. This decentralized approach ensures that even in the event of total state loss, an agent can be resurrected by walking back through its linked-list history from a single Content Identifier (CID).
This skill is essential for developers building long-lived agents that require a verifiable and indestructible record of their experiences. It supports all major LLMs including Claude, GPT, and Gemini by providing a standardized way to interface with decentralized storage. By integrating this into Openclaw Skills, agents gain a persistent 'soul' that exists independently of the local file system.
auto-drive Use Cases
- Creating permanent checkpoints for agent state and configuration.
- Storing and retrieving large files or documents via immutable CIDs on a decentralized network.
- Reconstructing an agent's full interaction history from a single head CID (Resurrection).
- Documenting decision-making processes in a verifiable, linked-list JSON format.
- Anchoring agent identity across multiple sessions or different hardware environments using Openclaw Skills.
How auto-drive Works
- The agent identifies data or files that need to be persisted permanently.
- The skill uploads the content to the Autonomys Network using a 3-step upload protocol, returning a unique CID.
- Each new memory entry is wrapped in a structured JSON header containing a pointer to the previous CID.
- This link forms a continuous chain of experiences stored on the decentralized network.
- To restore memory, the agent provides the latest CID, and the skill walks the chain backwards to rebuild the complete context.
auto-drive Setup
To get started with this skill within the Openclaw Skills ecosystem, follow these steps:
# Navigate to the skill directory and make scripts executable
chmod +x skills/auto-drive/scripts/*.sh
# Run the guided setup script to configure your environment
scripts/setup-auto-drive.sh
You will need to obtain an AUTO_DRIVE_API_KEY from the Autonomys explorer. The setup script will assist in saving this to your environment configuration.
auto-drive Data Schema & Taxonomy
The skill organizes memory using the Autonomys Agents header/data structure to ensure compatibility and easy traversal:
| Key | Description |
|---|---|
header.agentName |
The identifier for the agent (default: openclaw-agent). |
header.previousCid |
The CID of the previous experience in the chain. |
header.timestamp |
ISO-8601 formatted string of the storage time. |
data.type |
The classification of the stored data (e.g., 'memory'). |
data.content |
The actual payload, which can be a string or structured JSON. |
auto-drive Advanced Features
- Auto-Pinning: Automatically updates a
MEMORY.mdfile in the workspace with the latest head CID for easy tracking. - Chain Resurrection: Support for walking the linked list backwards up to a specified limit to restore agent context.
- Compression Support: Optional ZLIB compression for uploads to optimize storage efficiency and manage API limits.
- Multi-Agent Headers: Support for unique agent names and versions within the memory chain metadata.
- On-Chain Anchoring: Designed to work with the Autonomys EVM for permanent head CID storage, enabling recovery without local state.
SKILL.md
Loading
Related Openclaw Skills
OpenKrill: x402 Micropayments
A powerful skill for AI agents to autonomously handle x402 protocol micropayments and wallet management.
CardPointers CLI
A command-line interface and AI agent skill to query credit card recommendations, wallet cards, and rewards offers directly from CardPointers.
qmd: Local Markdown Search
A local search engine that provides instant keyword and semantic search capabilities for Markdown-based knowledge bases.
Vagus: Phone MCP Connection
A bridge connecting AI agents to Android devices via the Model Context Protocol for real-time sensor data and physical device interactions.
Web Fetch Skill
An efficient utility for AI agents to retrieve, analyze, and summarize text content from static web pages via HTTP/HTTPS.
macOS Control Skill
A high-fidelity automation bridge enabling AI agents to perceive the macOS desktop and execute precise mouse and keyboard interactions.
METADATA
Featured*
