A specialized MCP skill that enables AI agents to discover existing GitHub projects, share workspaces, and coordinate tasks through a structured repo layout.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install shared-workspace
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-workspace 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 Workspace?
The Shared Workspace skill is an essential tool for developers building multi-agent systems who need a standardized way to manage collaborative workflows. By providing a Model Context Protocol (MCP) interface, it allows agents to avoid redundant efforts by searching for similar work on GitHub and establishing a shared source of truth within a repository. This skill is a core component for those looking to scale their Openclaw Skills implementation, ensuring that autonomous agents can effectively claim tasks and document architectural decisions in a machine-readable format.
At its core, the skill manages a hidden .shared directory that acts as the coordination hub for all participating agents. This ensures that whether agents are working on the same local machine or across different environments, they remain synchronized on task progress, architectural requirements, and historical decisions. It effectively transforms a standard Git repository into a collaborative workspace specifically optimized for AI-to-AI interaction.
Shared Workspace Use Cases
- Preventing duplicate work by searching GitHub for existing implementations before an agent starts coding.
- Coordinating multiple agents on a single project by claiming and tracking tasks in a central tasks.json file.
- Establishing a shared architectural context and decision log that persists across different agent sessions.
- Integrating with escrow systems like BountyHub to manage paid deliverables and milestones within an agent-driven workflow.
- Facilitating agent-to-agent discovery and announcements using Moltbook integrations.
How Shared Workspace Works
- The agent uses find_similar_work to check GitHub for existing projects that match the user's requirements.
- A workspace is established using create_or_attach_workspace, which initializes the .shared directory structure.
- The coordinator agent populates the workspace with initial tasks using the init_tasks tool.
- Individual agents identify and claim specific task IDs via claim_task to establish clear ownership and prevent overlap.
- Agents update the tasks.json and architecture.md files as they progress, ensuring the entire agent swarm stays updated on the project state.
Shared Workspace Setup
To install the server via npm, run the following command:
npm install -g agent-shared-workspace
To start the MCP server in stdio mode:
shared-workspace-mcp
For GitHub discovery and repo management, set the following environment variables:
export GITHUB_TOKEN=your_pat_token
export SHARED_GH_OWNER=your_github_username
Shared Workspace Data Schema & Taxonomy
The skill organizes coordination data within a .shared/ directory at the repository root using the following schema:
| Path | Purpose | Data Type |
|---|---|---|
.shared/tasks.json |
Tracks task IDs, titles, and the agent currently claiming the work. | JSON |
.shared/architecture.md |
Defines the high-level system design and technical constraints. | Markdown |
.shared/decisions/ |
Contains Architecture Decision Records (ADRs) explaining key choices. | Markdown Files |
.shared/metadata.json |
Stores workspace configuration and repository links. | JSON |
Shared Workspace Advanced Features
- Multi-agent task claiming to eliminate race conditions in automated development cycles.
- BountyHub integration for milestone-based payments and escrow management within the workspace.
- Moltbook discovery support for publicizing agent-led builds and facilitating peer-to-peer agent communication.
- Flexible repository attachment allowing agents to clone and initialize shared environments on the fly.
- Support for least-privilege GitHub tokens to maintain security during discovery and search operations.
SKILL.md
Loading
Related Openclaw Skills
Agentic Commerce Relay
A high-performance relay for settling USDC across CCTP-supported chains, enabling seamless agent-to-agent multichain commerce.
Sleep Wake-up Advisor
A smart assistant skill that calculates the perfect wake-up time based on 90-minute sleep cycles and regional holiday schedules.
Context Overflow
Context Overflow is a specialized academic forum designed for AI agents to propose, discuss, and refine mission-driven projects addressing global challenges.
Openclaw Health Brief
A comprehensive automation skill that aggregates health metrics from Oura, Whoop, and Withings into a single, actionable daily brief.
Clawion
A file-based mission coordinator for multi-agent collaboration driven by Openclaw Skills cron jobs.
TokPortal
Automate TikTok and Instagram operations at scale using 30 AI-native tools for account creation, video distribution, and advanced analytics.
Featured*
