Co-operate with AI. The open-source platform where humans and AI run businesses together -- from strategy to execution.
DeskClaw is the operating platform for human-AI co-managed organizations. Through Cyber Workspaces, humans and AI operate as partners in a shared digital space -- humans provide strategic judgment, AI delivers relentless execution, and together they build something neither could alone.
We believe the future belongs to organizations where humans and AI co-operate -- not as master and tool, but as partners who each bring irreplaceable value to the business.
- Human operators bring strategic judgment, creative decisions, and value alignment -- deciding what to do and why
- AI operators bring tireless execution, pattern recognition, and rapid iteration -- pushing how to do it to the extreme
- Cyber Workspace is where co-operating happens -- a shared operations board (blackboard), task delegation, and real-time coordination that fuses human and AI capabilities into one
The digital space where humans and AI co-operate. Hexagonal topology visualizes your operating team's relationships; the shared blackboard serves as the team's operations dashboard; task publishing lets any partner -- human or AI -- delegate work to whoever is best suited. Not a monitoring panel, but the place where business happens.
Investment in AI operating capabilities. Loading a new Gene onto an AI partner opens a new dimension for your business -- modular capability packages from a public marketplace or your private enterprise library, composable on demand, continuously evolving. The business you run determines the Genes you load.
Instant expansion of operating capacity. One-click deployment of AI operating partners on Kubernetes clusters. Local development supported via dev.sh for rapid iteration.
- Cyber Workspace -- Hexagonal topology space where human and AI partners co-operate, share an operations board, and delegate tasks
- Gene System -- Modular capability investment: load new business dimensions onto AI partners from a public or private marketplace
- One-Click Scale -- Expand your operating capacity end-to-end, with SSE real-time progress streaming
- Multi-Cluster Operations -- Cross-cluster orchestration, health checks, and elastic scaling across your business footprint
graph TD
Human["Human Operators"] --> Portal["Portal"]
Portal --> API["Backend API Hub<br>Python 3.12 + FastAPI"]
API --> DB[(PostgreSQL)]
API --> CW["Cyber Workspace<br>Blackboard / Topology / Delegation"]
API --> Gene["Gene System<br>Public Marketplace"]
API --> Compute["K8s"]
Compute --> Runtime["AI Runtime<br>OpenClaw / Nanobot"]
Runtime <-->|"Channel Plugin (SSE)"| API
Runtime --> LLM["LLM Proxy"] --> Providers["OpenAI / Anthropic / Gemini / ..."]
AIPartners["AI Operating Partners"] -.->|"co-operate"| CW
DeskClaw/
├── nodeskclaw-portal/ # User Portal -- Vue 3 + Tailwind CSS
├── nodeskclaw-backend/ # API Server -- Python 3.12 + FastAPI + SQLAlchemy
├── nodeskclaw-llm-proxy/ # LLM Proxy -- Python + FastAPI
├── nodeskclaw-artifacts/ # Docker images & deploy manifests
├── openclaw-channel-nodeskclaw/ # Cyber Workspace channel plugin
├── openclaw-channel-dingtalk/ # DingTalk channel plugin (Stream protocol)
├── openclaw/ # DeskClaw runtime source (external)
└── vibecraft/ # VibeCraft source (external)
Full-stack internationalization covering Portal and Backend.
- Language detection:
zh*->zh-CN,en*->en-US, fallbacken-US - Error display: prefer
message_keylocal translation, fall back tomessagewhen missing - Backend contract:
code+error_code+message_key+message+data
DeskClaw is designed to run on Kubernetes. K8s is the primary deployment target for both staging and production. For local development, use dev.sh instead (see below).
Requires a K8s cluster, a container registry, and an external PostgreSQL database.
| Dependency | |
|---|---|
| Kubernetes cluster | 1.24+ with Ingress Controller (e.g. ingress-nginx) |
| Container registry | Any Docker V2 registry (Docker Hub, AWS ECR, GCR, etc.) |
| PostgreSQL | External database (e.g. AWS RDS, GCP Cloud SQL) |
| kubectl | Configured with access to your cluster |
| Docker | For building images locally |
# Create deploy/.env.local (git-ignored)
cat > deploy/.env.local <<'EOF'
REGISTRY="your-registry.example.com/deskclaw"
KUBE_CONTEXT="your-kubectl-context"
EOF
# Login to your container registry
docker login your-registry.example.comcp nodeskclaw-backend/.env.example nodeskclaw-backend/.env
# Edit .env -- fill in DATABASE_URL, JWT_SECRET, ENCRYPTION_KEY, etc.
# Minimum required:
# DATABASE_URL=postgresql+asyncpg://user:pass@your-rds:5432/nodeskclaw
# JWT_SECRET=<random-secret>
# ENCRYPTION_KEY=<32-byte-base64-key>Creates the namespace, uploads .env as a K8s Secret, and applies base Deployment + Service manifests:
./deploy/cli.sh init # Default: staging namespace
./deploy/cli.sh init --prod # Production namespace./deploy/cli.sh deploy # Build all images + rolling update (staging)
./deploy/cli.sh deploy --prod # Deploy to production (interactive confirm)
./deploy/cli.sh deploy backend # Deploy a single componentEdit deploy/k8s/ingress.yaml -- replace example.com hosts with your actual domains, then apply:
kubectl --context <CTX> -n <NS> apply -f deploy/k8s/ingress.yamlThe Ingress defines two hosts (configure as needed):
| Ingress | Default host | Backend service |
|---|---|---|
| Portal | console.example.com |
portal (80) + backend API (8000) |
| LLM Proxy | llm-proxy.example.com |
llm-proxy (80) |
See deploy/README.md for full CLI reference, image tagging, and the release/promote workflow.
| Dependency | |
|---|---|
| Python >= 3.12 + uv | Backend runtime & package manager |
| Node.js >= 18 + npm | Frontend runtime |
| PostgreSQL | Database (or use --docker-pg below) |
cd nodeskclaw-backend
cp .env.example .env
# Edit .env -- fill in DATABASE_URL, JWT_SECRET, etc../dev.sh # Start all services (backend + portal)
./dev.sh --docker-pg # Start a Docker PostgreSQL (no local PG install needed)
./dev.sh --fresh # Force reinstall all dependenciesThe script handles dependency installation, starts all services with colored log prefixes, and cleans up on Ctrl+C. --docker-pg launches a local PostgreSQL container automatically.
Services: backend (4510) + llm-proxy (4511) + portal (4517)
Manual Start (alternative)
Backend:
cd nodeskclaw-backend
uv sync
uv run uvicorn app.main:app --reload --port 4510API at http://localhost:4510 | Swagger at http://localhost:4510/docs | Auto-migration on first boot.
Frontend (Portal):
cd nodeskclaw-portal
npm install && npm run devPortal at http://localhost:4517 | /api auto-proxy to backend.
On first startup the backend prints the initial admin credentials directly in the terminal output:
========================================
Initial admin account
Account: admin
Password: <random>
Please change your password after login
========================================
Open http://localhost:4517 and sign in with the printed credentials. You will be prompted to change the password on first login.
K8s deployments are managed by deploy/cli.sh. The typical workflow is deploy to staging first, then promote to production.
Staging -- build images, push to registry, and rolling-update the staging namespace:
./deploy/cli.sh deploy --tag v0.9.0Production -- reuse the already-pushed images and update the production namespace (no rebuild):
./deploy/cli.sh promote v0.9.0Database migrations run automatically when the new backend pod starts. See deploy/README.md for full CLI usage and options.
- Back up your database before any major version upgrade.
- Check GitHub Releases for release notes and breaking changes.
- If your database was not previously managed by Alembic, you may need to run
alembic stamp headonce before upgrading. See Backend README for details.
| Backend | API hub, directory layout, env vars |
| Portal | User portal frontend |
| Artifacts | DeskClaw image build & deploy manifests |
| Channel Plugin | Cyber Workspace communication infrastructure |
| DingTalk Plugin | DingTalk channel via Stream protocol |
| LLM Proxy | AI reasoning capability relay |
- Discord -- Join the discussion, ask questions, share feedback
- GitHub Issues -- Bug reports and feature requests
- WeChat -- Scan the QR code below to join the developer group; if the WeChat group is full, please use Discord above
PRs welcome. See CONTRIBUTING.md for guidelines.