Free Trial
AI Applications

Getting Started with NemoClaw on Bitdeer AI Cloud

Johan Sim

• 4 min read
Getting Started with NemoClaw on Bitdeer AI Cloud

Deploy secure, production-ready AI agents in minutes with NVIDIA NemoClaw on Bitdeer AI Cloud. In this guide, you’ll learn how to quickly set up a fully sandboxed environment and run powerful models like Moonshot AI’s Kimi K2.5 through Bitdeer AI’s high-performance inference endpoints.

What You'll Build

By the end of this guide, you'll have:

  • A secure, sandboxed AI agent running on Bitdeer AI Cloud
  • Access to powerful models like Kimi K2.5 through Bitdeer AI's inference API
  • Full network policy controls and monitoring capabilities

Why NemoClaw on Bitdeer AI?

NemoClaw on Bitdeer AI Cloud brings together secure agent execution with scalable infrastructure and high-performance inference. While sandboxed execution with Landlock, seccomp, and network isolation is handled within the NemoClaw and OpenShell framework, Bitdeer AI Cloud can provide the underlying compute environment, enabling on-demand scaling of CPU and GPU resources with usage-based pricing. Our platform also delivers access to optimized inference endpoints for running leading models, along with network-level controls to help manage how agents interact with external services, ensuring both flexibility and operational control.

Prerequisites

NemoClaw System Requirements Minimum: 4+ vCPU, 16 GB RAM, 40 GB of free disk space, and Ubuntu 22.04 LTS.

One Bitdeer AI Cloud Requirements

Bitdeer AI Instance: Select g4a.xlarge (4 vCPU, 16 GB RAM)

Important: The sandbox image is approximately 2.4 GB compressed. For instances with 16 GB RAM, you'll have sufficient memory for smooth operation.

Software Dependencies

  • Node.js 20 or later
  • npm 10 or later
  • Docker (installed and running)
  • OpenShell CLI (installed by NemoClaw)

1.1 Create Virtual Machine

1. Log in to your Bitdeer AI Cloud account

2. Navigate to Virtual Machine services

3. Click Create Instance and select:

Instance Type: g4a.xlarge (4 vCPU, 16 GB RAM)

Storage: 40 GB SSD minimum

OS: Ubuntu 22.04 LTS

Region: Choose nearest to your location

4. Complete the order and wait for instance provisioning

5. Once running, note the public IP address

1.2 Access Your Instance

Connect via SSH:

ssh ubuntu@<your-instance-ip>

Step 2: Install Docker

NemoClaw requires Docker as the container runtime:

# Update package index
sudo apt-get update

# Install Docker (service starts automatically)
sudo apt-get install -y docker.io

# Add user to docker group
sudo usermod -aG docker $USER

Important: Log out and log back in for group changes to take effect:

exit
ssh ubuntu@<your-instance-ip>

Verify Docker installation:

docker --version
docker run hello-world

Step 3: Get Bitdeer AI API Key

Before installing NemoClaw, obtain your Bitdeer AI API key:

1. Log in to Bitdeer AI Cloud

2. Navigate to Models → API Keys

3. Click Generate API Key

4. Copy the generated key and save it securely

⚠ Security Warning: Never commit API keys to version control or share them publicly.

Step 4: Install and Configure NemoClaw

Download and run the NemoClaw installer, which includes an interactive wizard for configuring your Bitdeer AI inference endpoint:

curl -fsSL https://www.nvidia.com/nemoclaw.sh | bash

During the Installation Wizard

When prompted, configure Bitdeer AI as your inference provider:

[2/7] Configuring inference (NIM)
──────────────────────────────────────────────────

Inference options:
  1) NVIDIA Endpoints (recommended)
  2) OpenAI
  3) Other OpenAI-compatible endpoint  ← Select this
  4) Anthropic
  5) Other Anthropic-compatible endpoint
  6) Google Gemini

Choose [1]: 3

OpenAI-compatible base URL: https://api-inference.bitdeer.ai/v1
Other OpenAI-compatible endpoint API key: <paste-your-bitdeer-api-key>

Other OpenAI-compatible endpoint model []: nvidia/NVIDIA-Nemotron-3-Super-120B-A12B
# or: moonshotai/Kimi-K2.5

The wizard will:

  • Install Node.js 20+ if not present
  • Install npm packages
  • Set up the NemoClaw CLI
  • Configure Bitdeer AI as your inference provider
  • Save your API key securely to ~/.nemoclaw/credentials.json
  • Create your first sandbox

Post-Installation Summary

After installation completes, you should see:

──────────────────────────────────────────────────
Sandbox      my-bitdeer-agent (Landlock + seccomp + netns)
Model        nvidia/NVIDIA-Nemotron-3-Super-120B-A12B (Bitdeer AI Endpoint)
──────────────────────────────────────────────────
Run:         nemoclaw my-bitdeer-agent connect
Status:      nemoclaw my-bitdeer-agent status
Logs:        nemoclaw my-bitdeer-agent logs --follow
──────────────────────────────────────────────────

[INFO]  === Installation complete ===

If nemoclaw is not found after installation, run:

source ~/.bashrc
# or for zsh users:
source ~/.zshrc

Verify installation:

nemoclaw --version

Step 5: Connect and Chat with Your Agent

5.1 Connect to the Sandbox

nemoclaw my-bitdeer-agent connect

You'll enter the sandbox shell:

sandbox@my-bitdeer-agent:~$

5.2 Interactive TUI Mode

Launch the OpenClaw TUI for interactive chat:

openclaw tui

Send test messages and verify responses.

5.3 CLI Mode

For programmatic access or long outputs, use the CLI:

openclaw agent --agent main --local -m "Hello, how can you help me today?" --session-id test

To exit the sandbox shell:

exit

What's Next?

Now that you have your NemoClaw agent running on Bitdeer AI Cloud, here are some ways to extend it:

  • Connect to Telegram

Enable mobile access to your agent:

nemoclaw my-bitdeer-agent telegram-setup
  • Try Different Models

Experiment with other available models through Bitdeer AI Model Studio:

openshell inference set --provider openai --model zai-org/GLM-5
  • Monitor Activity

Launch the OpenShell TUI to watch your agent in action:

openshell tui
  • Customize Security

Define custom network policies to control what your agent can access.

Conclusion

With NemoClaw running on Bitdeer AI Cloud, deploying secure, production-ready AI agents becomes both practical and scalable. By combining NemoClaw’s sandboxed execution and policy framework with Bitdeer AI’s flexible infrastructure and optimized inference endpoints, you can move from setup to real-world usage in just a few steps. Whether you are experimenting with AI agents or building toward production, Bitdeer AI Cloud offers a reliable environment to run, scale, and manage your workloads with confidence.

Resources

Documentation:

Community & Support: