Yuan-lab-LLM/ClawManager

What it solves

ClawManager addresses the complexity of deploying and managing multiple AI agent instances at scale. It provides a centralized control plane to handle runtime orchestration, governed AI model access, resource sharing, and security for various agent frameworks without requiring users to manage the underlying Kubernetes infrastructure manually.

How it works

Built as a Kubernetes-native platform, ClawManager acts as an orchestration layer between administrators, users, and AI runtimes. It uses a Go backend and React frontend to manage the lifecycle of agent workspaces (in either shared "Lite" or dedicated "Pro" modes). AI traffic is routed through a governed AI Gateway for auditing and cost control, while an Agent Control Plane synchronizes the desired state and dispatches commands to the instances. It also integrates a Redis Team Bus for collaborative multi-agent workflows.

Who it’s for

• Platform teams responsible for hosting AI agent instances for multiple users.
• Operators who need visibility into runtime health, command dispatch, and state control.
• AI developers who want governed model access and reusable resource injection (like skills and channels) instead of manual per-instance setup.

Highlights

• Multi-Runtime Support: Manages workspaces for OpenClaw, Hermes, OpenCode, and DeepSeek Harness.
• AI Gateway: Centralized entry point for models with built-in cost accounting, risk rules, and end-to-end audit traces.
• Team Collaboration: Leader-mediated workflows with task decomposition, an Execution Kanban, and shared artifacts.
• Security Protection Platform: A comprehensive admin console for runtime defense, container isolation, and emergency circuit breakers.
• Resource Management: A Skill Hub and configuration center for reusable startup content and resource packs.