A powerful orchestration framework for spawning and managing multiple coordinated AI sub-agents to solve complex, multi-stage tasks in parallel.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install fuzzy-multi-agent-team
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 fuzzy-multi-agent-team 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 Multi-Agent Team Orchestration?
The Multi-Agent Team skill transforms a single AI into a sophisticated orchestrator capable of spawning, coordinating, and managing a fleet of specialized sub-agents. By leveraging this Openclaw Skills capability, developers can break down monolithic tasks into manageable sub-tasks handled by dedicated workers, such as researchers, coders, and reviewers. This modular approach significantly increases processing speed and improves the quality of complex outputs through specialized focus.
This skill supports various architectural patterns, from one-shot disposable teams to persistent squads that maintain long-term context. Whether you are running a competitive agent council for decision-making or a hierarchical tree for large-scale content generation, Multi-Agent Team provides the necessary primitives for communication, result synthesis, and lifecycle management within the Openclaw Skills ecosystem.
Multi-Agent Team Orchestration Use Cases
- Parallelized web research and data processing across multiple concurrent workers.
- Creating an agent council or debate to evaluate complex decisions from multiple perspectives.
- Decomposing large-scale software engineering tasks into parallel coding and review sub-tasks.
- Orchestrating sequential pipelines where the output of a research agent feeds an analysis agent and then a writer agent.
- Managing persistent squads of specialized AI workers for ongoing project collaboration.
How Multi-Agent Team Orchestration Works
- The orchestrator receives a complex task and decomposes it into independent, actionable sub-tasks.
- Specialized sub-agents are spawned using the sessions_spawn command with specific roles and instructions.
- The orchestrator manages the lifecycle, choosing between one-shot execution or persistent sessions for ongoing work.
- Results are gathered either through automatic announcements, blocking waits via sessions_yield, or manual history polling.
- The orchestrator synthesizes the disparate data points from all sub-agents into a single, high-quality final response.
- Stale agents are identified and terminated to optimize resource usage and system performance.
Multi-Agent Team Orchestration Setup
To begin using Multi-Agent Team within your Openclaw Skills environment, ensure your primary agent has the orchestration capability enabled. You can initiate sub-agents directly through command execution.
# Example command to spawn a one-shot research agent
sessions_spawn(task="Research Openclaw Skills architecture", runtime="subagent", mode="run")
Configure timeouts and session modes based on your task complexity to ensure efficient token usage and system responsiveness.
Multi-Agent Team Orchestration Data Schema & Taxonomy
The skill organizes data through unique session keys and shared workspace files to maintain state across the team.
| Data Component | Format | Description |
|---|---|---|
| Session Key | String | Unique identifier for each spawned sub-agent. |
| Runtime | Enum | Set to subagent for internal Openclaw orchestration. |
| Shared Workspace | Files | Common file directory (e.g., /workspace/) for inter-agent state sharing. |
| Session History | JSON Array | Log of messages and results produced by a specific agent session. |
| Metadata | Object | Includes runTimeoutSeconds, activeMinutes, and agent kinds. |
Multi-Agent Team Orchestration Advanced Features
- Support for Fan-Out and Fan-In patterns to handle massive parallel data processing and synthesis.
- Hierarchical agent trees allowing sub-agents to spawn their own nested workers for deep task recursion.
- Persistent squad management where agents maintain long-term memory across multiple interactions.
- Integrated cleanup tools to monitor and kill active sub-agent sessions based on inactivity or completion.
- Flexible communication hooks including blocking yield and asynchronous history polling for complex orchestration.
SKILL.md
Loading
Related Openclaw Skills
Enterprise Agent Commerce Playbook
A comprehensive organizational and technical guide for Fortune 500 companies to adopt autonomous agent commerce for B2B transactions.
XHS Big Text Post Creator
A comprehensive AI workflow to generate visually striking Little Red Book posts with optimized text overlays and automated delivery.
Greenhelix Formal Gatekeeper
A formal verification proxy that uses the Z3 SMT solver to mathematically prove the safety of autonomous agent plans before execution.
Desktop AI Agent
A powerful automation engine that enables AI agents to control desktop environments through vision-based navigation and keyboard/mouse emulation.
Computer Takeover
A powerful remote orchestration skill for full unattended control of paired Windows, macOS, and mobile devices.
Eight-Step Idea to Reality Method
A systematic execution framework designed to bridge the gap between abstract concepts and concrete reality by integrating human psychology and risk management.
Featured*
