/README.md

ClawManager

A Kubernetes-first control plane for managing OpenClaw and Linux desktop runtimes at team and cluster scale.

Languages: English | 中文 | 日本語 | 한국어 | Deutsch

News

• [2026-03-26]: 🚀🚀 AI Gateway documentation and overview were refreshed, including model governance, audit and trace, cost accounting, and risk control. See AI Gateway.
• [2026-03-20]: 🎉🎉 ClawManager Release — ClawManager is now a virtual desktop management platform featuring batch deployment, Webtop support, desktop portal access, runtime image settings, OpenClaw memory/preferences Markdown backup and migration, cluster resource overview, and multilingual documentation.

What It Is

ClawManager helps teams deploy, operate, and access desktop runtimes on Kubernetes from one place.

It is built for environments where you need to:

• create desktop instances for multiple users
• control quotas, runtime images, and lifecycle centrally
• keep desktop services inside the cluster
• give users secure browser access without exposing pods directly

Why Users Pick It

• One admin panel for users, quotas, instances, and runtime images
• OpenClaw support with import/export for memory and preferences
• Secure desktop access through the platform instead of direct pod exposure
• AI Gateway governance for controlled model access, audit trails, cost analysis, and risk controls
• Kubernetes-native deployment and operations flow
• Works for both admin-managed rollout and self-service instance creation

Quick Start

Prerequisites

• A working Kubernetes cluster
• kubectl get nodes works

Deploy

Apply the bundled manifest:

kubectl apply -f deployments/k8s/clawmanager.yaml
kubectl get pods -A
kubectl get svc -A

Build From Source

If you want to run or package ClawManager from source instead of using the bundled Kubernetes manifest:

Frontend

cd frontend
npm install
npm run build

Backend

cd backend
go mod tidy
go build -o bin/clawreef cmd/server/main.go

Docker Image

Build the full application image from the repository root:

docker build -t clawmanager:latest .

Default Accounts

• Default admin account: admin / admin123
• Default password for imported admin users: admin123
• Default password for imported regular users: user123

First Use

1. Log in as admin.
2. Create or import users and assign quotas.
3. Review or update runtime image cards in system settings.
4. Log in as a user and create an instance.
5. Access the desktop through Portal View or Desktop Access.

Main Capabilities

• Instance lifecycle management: create, start, stop, restart, delete, inspect, and sync
• Runtime types: openclaw, webtop, ubuntu, debian, centos, custom
• Runtime image card management from the admin panel
• User quota control for CPU, memory, storage, GPU, and instance count
• Cluster resource overview for nodes, CPU, memory, and storage
• Token-based desktop access with WebSocket forwarding
• AI Gateway for model management, traceable audit logs, cost accounting, and risk control
• CSV-based bulk user import
• Multilingual interface

AI Gateway

AI Gateway is the governance plane for model access inside ClawManager. It gives OpenClaw instances a single OpenAI-compatible entry point while adding policy, audit, and cost controls on top of upstream providers.

• Model management for regular and secure models, provider onboarding, activation, endpoint configuration, and pricing policy
• End-to-end audit and trace records for requests, responses, routing decisions, and risk hits
• Built-in cost accounting with token tracking and estimated usage analysis
• Risk control with configurable rules and automated actions such as block and route_secure_model

Supported Model Service Platforms

ClawManager includes built-in vendor templates for:

• OpenAI
• OpenRouter
• DeepSeek
• SiliconFlow
• Moonshot AI
• Zhipu AI
• Alibaba DashScope
• Volcengine Ark
• Groq
• Together AI
• Fireworks AI
• xAI
• Perplexity
• 01.AI
• MiniMax
• Local / Internal endpoints

Local / Internal can also be used for self-hosted OpenAI-compatible gateways, Ollama, One API, and other private model endpoints.

For screenshots, the full feature breakdown, and the model selection and routing flow, see docs/aigateway.md.

Product Flow

1. An admin defines users, quotas, and runtime image policies.
2. A user creates an OpenClaw or Linux desktop instance.
3. ClawManager creates and tracks the Kubernetes resources.
4. The user accesses the desktop through the platform.
5. Admins monitor health and capacity from the dashboard.

Architecture

Browser
  -> ClawManager Frontend
  -> ClawManager Backend
  -> MySQL
  -> Kubernetes API
  -> Pod / PVC / Service
  -> OpenClaw / Webtop / Linux Desktop Runtime

Configuration Notes

• Instance services stay on Kubernetes internal networking
• Desktop access goes through the authenticated backend proxy
• Runtime images can be overridden from system settings
• Backend deployment is best kept inside the cluster

Common backend environment variables:

• SERVER_ADDRESS
• SERVER_MODE
• DB_HOST
• DB_PORT
• DB_USER
• DB_PASSWORD
• DB_NAME
• JWT_SECRET

CSV Import Template

Username,Email,Role,Max Instances,Max CPU Cores,Max Memory (GB),Max Storage (GB),Max GPU Count (optional)

Notes:

• Email is optional
• Max GPU Count (optional) is optional
• all other columns are required

License

This project is licensed under the MIT License.

Open Source

Issues and pull requests are welcome.