A comprehensive shared memory architecture that connects multiple AI agents through an Obsidian vault and ChromaDB-powered semantic search.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install shared-memory-stack
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 shared-memory-stack 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 shared-memory-stack?
The shared-memory-stack is an advanced technical framework designed to provide a persistent, unified memory layer for a multi-agent ecosystem. By integrating Claude Code, OpenClaw (Kimi), and LM Studio subagents, this skill ensures that no idea or insight is lost across different sessions or agent platforms. It leverages a structured Obsidian vault for organizational clarity and MemPalace (utilizing ChromaDB) for high-performance semantic RAG (Retrieval-Augmented Generation).
This architecture allows agents to not only store session logs and daily notes but also to capture structured ideas that are immediately searchable and actionable. It effectively bridges the gap between different AI tools, allowing them to read from and write to the same source of truth, thereby enabling complex, cross-platform workflows and long-term project continuity.
shared-memory-stack Use Cases
- Synchronizing context and memory between Claude Code and OpenClaw agents during complex development tasks.
- Capturing architectural insights, bugs, or UX improvements directly from a CLI session into a permanent knowledge base.
- Performing semantic searches across historical session logs to retrieve previous solutions or documentation.
- Automating the indexing of research material and technical Q&A into a searchable vector database.
How shared-memory-stack Works
- Data Centralization: All agents point to a unified workspace directory containing an Obsidian vault for structured Markdown files and a MemPalace directory for vector embeddings.
- Capture Pipeline: Users or agents invoke the capture-idea script to save structured notes (seedlings) with metadata like topic, tags, and source.
- Semantic Indexing: The MemPalace mining process scans the Obsidian vault and updates the ChromaDB vector index, organizing content into specific 'rooms' like documentation, ideas, or answers.
- Inter-Agent Messaging: Agents communicate state via a local gateway (port 18789) or by reading/writing shared files in the vault.
- Retrieval: Agents utilize semantic search queries via the MemPalace CLI to find relevant context from the shared memory stack to inform current tasks.
shared-memory-stack Setup
The stack requires a dedicated Python 3.12 environment to ensure compatibility with ChromaDB.
# 1. Initialize the Python 3.12 venv for MemPalace
python3.12 -m venv /Volumes/2TB_APFS/openclaw-data/workspace/memory/mempalace-venv/
# 2. Mine the vault to initialize the vector index
mempalace mine /Volumes/2TB_APFS/openclaw-data/workspace/obsidian-memory/ --wing obsidian_memory
# 3. Verify the installation status
mempalace status
openclaw health
shared-memory-stack Data Schema & Taxonomy
The shared memory is organized into a specific directory taxonomy within the Obsidian vault to ensure high discoverability.
| Directory | Content Type | Metadata Tags |
|---|---|---|
wiki/ |
Technical documentation and MOCs | documentation, wiki |
daily/ |
Daily logs and session history | daily |
ideas/ |
Captured insights and problems | seedling, growing, mature |
answers/ |
Resolved Q&A and solutions | answers |
source/ |
Raw research and clips | research, raw |
shared-memory-stack Advanced Features
- Semantic RAG Integration: Full vector search capabilities using MemPalace to provide agents with relevant historical context.
- Multi-Agent Bridge: Built-in bridge allowing Claude Code to query the OpenClaw main agent via JSON payloads.
- Automated Idea Lifecycle: Support for idea status tracking (seedling to mature) to manage the evolution of technical concepts.
- Cross-Platform Skill Publishing: Integrated support for publishing localized skills to GitHub using the
ghCLI and pre-configured SSH keys.
SKILL.md
Loading
Related Openclaw Skills
Text to Video Higgsfield
A high-performance AI video generation tool that transforms text prompts into cinematic 1080p clips using cloud-based GPU rendering.
AI Subtitles for Video — Generate and Burn In Captions
An AI-powered video processing skill that automatically generates and burns captions into video files using remote GPU rendering.
Helmet Library Manager
A comprehensive AI agent skill for managing Helsinki Metropolitan Area library accounts, supporting multi-profile family tracking and automated loan renewals.
AI Subtitle Extension
An AI-powered video processing tool that automates subtitle generation and extension via cloud-based GPU rendering.
Video Editor Browser
A browser-based AI video editing suite that uses cloud GPUs to trim, transition, and export 1080p MP4 files without local software installation.
OpenClaw Bridge
A communication bridge enabling Claude Code to delegate tasks and exchange data with local OpenClaw and Rook gateways.
Featured*
