OpenClaw Multi-Agent Setup: Separate Tools, Memory, and Shared Repo

In this LBSocial tutorial, discover how to build a cleaner, safer, and more efficient multi-agent system in OpenClaw. Instead of overloading a single AI agent with all tasks, this guide demonstrates how to architect the system and split it into three specialized agents running on a single Google Cloud VM: a Main Coordinator, a Data Specialist, and a Google Workspace Specialist. Learn how to configure the AGENTS.md, TOOLS.md, and SOUL.md files to assign specific skills and enforce strict guardrails. By giving each agent a separate workspace memory, restricting their tools, and allowing them to hand off work through a shared local Git repository, developers can create a multi-agent team that is highly secure and token-efficient. Finally, see how to route these agents through three separate Telegram bots for effortless mobile control. 🔗 Code & Resources: Access all the terminal commands, configuration scripts, and the exact .md files used in this video on the LBSocial website: https://www.lbsocial.net/post/openclaw-multi-agent-setup ⏱️ Timeline: 00:00 - Introduction: Why build a Multi-Agent System? 00:32 - Demo: Interacting with 3 Separate Telegram Bots 00:56 - Architecture Overview: OpenClaw VM and Shared Repo 02:19 - Creating the Shared Git Repository 02:38 - Adding Data and GWS Agents in the Terminal 03:29 - Passing Skills to Specific Agents 03:51 - Harness Engineering: Configuring AGENTS.md, TOOLS.md, and SOUL.md 06:41 - Reviewing the openclaw.json Configuration 07:11 - Setting up Multiple Telegram Bots (BotFather) 08:42 - Binding Agents to Telegram Bots via CLI 09:20 - Live Test: Data Agent performing Python Analysis 10:00 - Live Test: GWS Agent managing Google Drive & Classroom 10:34 - Conclusion: The Benefits of Separated Workflows 📺 Previous OpenClaw Tutorials in this Series: OpenClaw + Google Cloud: Build a 24/7 Personal AI Assistant https://www.youtube.com/watch?v=29uolWUiEvA OpenClaw + GitHub: Building a Real AI Teammate https://www.youtube.com/watch?v=yBbLs3VAayc OpenClaw Skills: Guiding AI Agents for Data Analytics https://www.youtube.com/watch?v=RLUa-j-MpgQ OpenClaw + Google Workspace CLI: Build an AI Agent for Gmail, Docs, and Classroom https://www.youtube.com/watch?v=_ssB1YXRRtk OpenClaw + VS Code SSH: Manage Your Google Cloud VM in VS Code https://www.youtube.com/watch?v=KH4rvjsK0mg #OpenClaw #AIAgents #MultiAgentSystem #LBSocial #AIAutomation #GoogleWorkspace