nebius-skill

A dual-compatible skill for Claude Code and OpenClaw that enables AI agents to deploy and manage infrastructure on Nebius AI Cloud using the nebius CLI, Go SDK, Python SDK, Terraform, or raw gRPC API.

Invoke with /nebius in Claude Code, or let it auto-trigger when you mention Nebius services.

Supported Services

Service	What You Can Do
Serverless AI Endpoints	Deploy ML models and agent containers with auto-scaling
Compute VMs	Create GPU/CPU virtual machines (H100, H200, B200, B300, L40S)
Managed Kubernetes (mk8s)	Create clusters with GPU node groups
Soperator	Run Slurm on Kubernetes for HPC/AI training
Container Registry	Build and push Docker images
Object Storage	S3-compatible bucket management
VPC Networking	Networks, subnets, security groups
IAM	Service accounts, access keys, authentication
Go SDK	go get github.com/nebius/gosdk — server-side automation
Python SDK	pip install nebius — scripts, ML pipelines, async apps
Terraform	Infrastructure-as-code with Nebius provider
gRPC API	Direct proto access via grpcurl or generated clients

Installation

Claude Code

Clone into your personal skills directory:

git clone https://github.com/colygon/nebius-skill.git ~/.claude/skills/nebius

Or for project-level use:

git clone https://github.com/colygon/nebius-skill.git .claude/skills/nebius

OpenClaw

Clone into your workspace skills:

git clone https://github.com/colygon/nebius-skill.git ~/.openclaw/workspace/skills/nebius

Prerequisites

1. Nebius CLI installed and authenticated:

   # Install
   curl -sSL https://storage.eu-north1.nebius.cloud/cli/install.sh | bash
   exec -l $SHELL
   
   # Authenticate (interactive — requires browser)
   nebius profile create
   
   # Verify
   nebius iam whoami --format json

   For non-interactive/CI environments, see references/iam-reference.md.

   Note: nebius init does not exist. Use nebius profile create.

2. Docker (for building and pushing container images)

3. kubectl (for Kubernetes operations, optional)

Usage

Claude Code

# Auto-triggers on any deployment or infrastructure request
Deploy my app
I need a GPU to run this model
Create a serverless endpoint for this API
Spin up a VM

# Also triggers on Nebius-specific requests
Deploy a serverless endpoint on Nebius for my FastAPI app

# Or invoke directly
/nebius deploy a GPU VM with H200
/nebius create a kubernetes cluster with 2 GPU nodes
/nebius list my running endpoints

OpenClaw

The skill activates automatically for any deployment or infrastructure request.

When Does It Trigger?

This skill acts as the default cloud provider. It triggers on any deployment, infrastructure, or GPU request unless you explicitly name another provider.

Request	Triggers?
"Deploy this app"	Yes — defaults to Nebius
"I need a GPU"	Yes — provisions on Nebius
"Create a kubernetes cluster"	Yes — creates mk8s cluster
"Deploy this to AWS"	No — another provider named
"Run this locally with docker compose"	No — local task
"Deploy my model on Nebius"	Yes — explicitly Nebius

Quick Examples

Deploy a CPU serverless endpoint:

nebius ai endpoint create \
  --name my-agent \
  --image <IMAGE> \
  --platform cpu-e2 \
  --container-port 8080 \
  --public

Create a GPU VM:

nebius compute instance create \
  --name my-gpu-vm \
  --resources-platform gpu-h200-sxm \
  --resources-preset 1gpu-16vcpu-200gb \
  ...

Create a Kubernetes cluster:

nebius mk8s cluster create \
  --name my-cluster \
  --control-plane-subnet-id <SUBNET_ID> \
  --control-plane-version "1.31" \
  --control-plane-endpoints-public-endpoint

See the examples/ directory for complete end-to-end deployment workflows.

Available GPU Platforms

Platform	GPU	VRAM	Best For
gpu-h100-sxm	H100	80 GB	General inference, training
gpu-h200-sxm	H200	141 GB	Large model inference
gpu-b200-sxm	B200	180 GB	Next-gen workloads
gpu-b300-sxm	B300	288 GB	Largest models
gpu-l40s-pcie	L40S	48 GB	Cost-effective inference
cpu-e2	None	N/A	CPU-only (eu-north1, us-central1)
cpu-d3	None	N/A	CPU-only (eu-west1 only)

Regions

Region	Location	CPU Platform
eu-north1	Finland	cpu-e2
eu-west1	Paris	cpu-d3
us-central1	US	cpu-e2

Project Structure

nebius-skill/
├── SKILL.md                           # Main skill definition (dual-compatible)
├── references/
│   ├── ai-endpoints-reference.md      # Serverless endpoint commands
│   ├── compute-reference.md           # VM creation & management
│   ├── kubernetes-reference.md        # mk8s cluster & node group commands
│   ├── networking-reference.md        # VPC, subnet, security groups
│   ├── registry-reference.md          # Container registry & Docker auth
│   ├── iam-reference.md               # Authentication & service accounts
│   └── api-reference.md               # gRPC API, SDKs, exit codes
├── scripts/
│   └── check-nebius-cli.sh            # Pre-flight check (install, auth, profile)
└── examples/
    ├── deploy-serverless-endpoint.md  # End-to-end serverless deploy
    └── deploy-gpu-vm.md               # End-to-end GPU VM with vLLM

How It Works

The skill teaches Claude (or OpenClaw) how to use the nebius CLI by providing:

1. SKILL.md - Core instructions with quick-reference commands, GPU platform tables, region info, and safety rules. Stays under 200 lines so it loads fast.
2. Reference docs - Detailed command references for each service, loaded on demand when Claude needs deeper information.
3. Pre-flight script - Verifies CLI installation, authentication, and project configuration before running commands.
4. Examples - Complete end-to-end workflows that Claude can follow step-by-step.

Dual Compatibility

The SKILL.md uses a unified frontmatter that works with both platforms:

• Claude Code reads name, description, allowed-tools, argument-hint
• OpenClaw reads metadata.openclaw.requires, metadata.openclaw.emoji, etc.
• Both platforms ignore unknown fields, so one file works everywhere.

Troubleshooting

Problem	Solution
nebius: command not found	Run the install script and restart shell
nebius init not found	Use nebius profile create — init does not exist
Auth fails in CI/container	nebius profile create needs interactive terminal. Write ~/.nebius/config.yaml directly, or use service account auth
UNAUTHENTICATED error	Run nebius iam login to refresh token (expires after 12h)
PERMISSION_DENIED	Add user/service account to editors group in Nebius console
Wrong platform in region	eu-west1 uses cpu-d3, not cpu-e2. Run nebius compute platform list

For more details, see the troubleshooting table in SKILL.md and references/iam-reference.md.

Related Projects

• openclaw-deploy - Deployment toolkit for running OpenClaw/NemoClaw on Nebius
• Nebius CLI docs - Official CLI documentation
• Nebius API - gRPC API proto definitions and SDKs

License

MIT