A production-tested architecture for building and orchestrating teams of 5-10 specialized AI agents with cross-agent routing.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install multi-agent-blueprint
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 multi-agent-blueprint 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 Blueprint?
The Multi-Agent Blueprint is a professional-grade template designed for developers looking to scale their automation using Openclaw Skills. It provides a structured approach to building teams of specialized agents—such as Coordinators, Finance Advisors, and Tech Leads—that can communicate and hand off tasks autonomously. This framework addresses the complexities of multi-agent orchestration by providing standardized role templates and communication protocols.
By leveraging the power of Openclaw Skills, this blueprint allows for sophisticated model tiering, ensuring that high-reasoning tasks are handled by premium models like Claude Opus while routine operations are routed to more economical tiers. This architecture is battle-tested for reliability, featuring fallback chains and centralized master patterns for file and database operations to prevent data conflicts and optimize resource usage.
Multi-Agent Blueprint Use Cases
- Creating a 24/7 autonomous business operation with specialized AI agents for finance, sales, and marketing.
- Scaling complex DevOps workflows where multiple Openclaw Skills collaborate to monitor infrastructure and manage file deployments.
- Implementing a centralized front-door coordinator that routes user requests to specialized backend agents via Telegram.
- Building resilient AI systems that utilize local models for heartbeats to maintain uptime without incurring high API costs.
How Multi-Agent Blueprint Works
- Team Planning: Select 3-10 specialized roles based on the tasks you need to automate, assigning each a specific model tier.
- Environment Setup: Create isolated workspace and agent directories for each member of the team to ensure data security and memory isolation.
- Core Configuration: Update the Openclaw Skills settings to enable agent-to-agent communication and define the model fallback order for maximum resilience.
- Persona Definition: Author the four mandatory Markdown files (IDENTITY, SOUL, AGENTS, and USER) to establish the agent's behavior and routing logic.
- Channel Integration: Connect each agent to Telegram or other messaging platforms using account-specific bot tokens and session scopes.
- Memory Persistence: Enable RAG and automated memory flushing to ensure context is preserved across session resets and system reboots.
Multi-Agent Blueprint Setup
To initiate the Multi-Agent Blueprint, first create the directory structure for your specialized agents:
# For each agent in your team:
mkdir -p ~/.openclaw/workspace-{agentname}/memory
mkdir -p ~/.openclaw/agents/{agentname}/agent
Then, add your agents to the agents.list in your openclaw.json and enable cross-agent communication:
{
"agentToAgent": { "enabled": true },
"agents": {
"list": [
{
"id": "finance",
"name": "finance",
"workspace": "~/.openclaw/workspace-finance",
"agentDir": "~/.openclaw/agents/finance/agent",
"model": "anthropic/claude-sonnet-4-5"
}
]
}
}
Ensure each agent directory contains the four required Markdown configuration files to define their operational logic within the Openclaw Skills ecosystem.
Multi-Agent Blueprint Data Schema & Taxonomy
The blueprint organizes data to support multi-agent autonomy and persistent memory storage:
| File/Directory | Role |
|---|---|
workspace-name/memory/ |
Contains the RAG database and long-term context files for the agent. |
agentDir/IDENTITY.md |
Defines the agent's name, role, and professional persona. |
agentDir/SOUL.md |
Detailed behavioral rules, expertise limits, and response length guidelines. |
agentDir/AGENTS.md |
The internal routing table for cross-agent communication using Openclaw Skills. |
agentDir/USER.md |
Local context regarding the user's business, timezone, and language preferences. |
Multi-Agent Blueprint Advanced Features
- Cross-agent routing using
sessions_sendwith unique session keys for targeted task handoffs. - Multi-tier model strategies that mix premium APIs with local Ollama heartbeats to slash operational costs.
- Comprehensive fallback chains that automatically switch providers if the primary model is rate-limited or offline.
- File Master and Notion Master patterns that centralize sensitive API keys and prevent race conditions in data writes.
- Memory search with hybrid BM25 and vector retrieval to maintain high context accuracy across large teams using Openclaw Skills.
SKILL.md
Loading
Related Openclaw Skills
Auto-Whisper Safe
A memory-efficient transcription tool that uses automatic audio chunking to run OpenAI Whisper on systems with limited RAM.
Agent Soul Crafter
A production-ready framework for designing structured AI agent personalities that feel authentic and follow strict response rules.
Agent Cost Monitor
A comprehensive monitoring skill to track token usage, costs, and optimization opportunities across all your AI agents.
Bolt Sprint Management
A comprehensive tool for managing software development sprints and stories within the Bolt platform using AI-driven automation.
News Article Generator
An automated tool that transforms Google search results into structured Markdown articles with downloaded image assets.
iOS View Generator
An AI-powered developer tool that converts UI screenshots into structured, lazy-loaded Objective-C code for iOS development.
Featured*
