Provider Help

Start earning with Mafutai

Ready to plug your GPUs into Africa’s AI supergrid? This guide covers prerequisites, software stack, and the steps to register, deploy the provider agent, and list your hardware.

Before you begin

Modern Linux host

Ubuntu 22.04/24.04 or Debian 12 with kernel drivers for NVIDIA or AMD GPUs. Ensure you have sudo access and that your GPUs support persistence mode.

Container runtime

Docker Engine 24+ or Podman with Docker compatibility. Mafutai’s provider agent ships as a container for predictable deployments.

Network & security

Outbound HTTPS (443) to api.mafutai.co.za, inbound SSH for maintenance, and the ability to mesh over WireGuard. Keep BIOS virtualization and IOMMU enabled.

Software you’ll need

GPU Drivers & Runtime
Install the latest NVIDIA Data Center driver (555.x) or AMD ROCm stack. Mafutai validates CUDA 12.5 and ROCm 6.x for inference and training workloads.
Container runtime
Docker Engine 24+ (or Podman) with GPU passthrough. Linux hosts use the NVIDIA Container Toolkit; Windows uses WSL2 + Docker Desktop with GPU support enabled.
Mafutai provider agent
Handles telemetry, heartbeats, and workload orchestration. You’ll deploy it via the Provider Portal once your account is approved.

Installation & registration

Request onboarding

Complete the Early Access form or email providers@mafutai.co.za. The team will schedule a short infrastructure validation call and provision your portal access.

Register provider account

Sign in to the Mafutai Provider Portal, submit KYC documents, and add your payout bank details. Verified providers receive a secure API key for the agent.

Deploy the agent

Run the installation script provided in the portal. It pulls the `mafutai/provider-agent` container, injects your API key, and registers each GPU node with its zone (e.g. JHB-01).

List your GPUs

Use the dashboard to describe each GPU (model, VRAM, power envelope). Set hourly ZAR pricing and choose availability windows (public, private, or enterprise-only).

Go live & monitor

Once the nodes report healthy telemetry, Mafutai approves the listing. Monitor rentals in the dashboard, review audit logs, and download Stripe-ready payout statements.

Example: install agent on Ubuntu

sudo apt update && sudo apt install -y docker.io nvidia-container-toolkit
sudo systemctl enable --now docker
sudo nvidia-ctk runtime configure --runtime=docker
sudo systemctl restart docker

curl -sSL https://portal.mafutai.co.za/agent.sh |   API_KEY=<your_api_key_here> HOSTNAME=JHB-01-A6000 bash

You’ll receive your personal install script and API key once the portal approves your provider account.

Frequently asked

Can I start with a single GPU?

Yes. Many providers begin with one RTX 6000 or A100. You can scale horizontally as demand grows.

What if my GPU goes offline?

The provider agent sends heartbeat alerts. Mafutai automatically pauses listings after repeated failures and helps you triage the node.

How do payouts work?

Usage-based billing flows through Stripe. Choose weekly or monthly settlements in the Provider Portal and receive EFT payouts in ZAR.

OS-specific setup

Linux hosts

1. Install drivers
Use the official NVIDIA packages (e.g. `sudo apt install nvidia-driver-555`) or AMD ROCm meta packages. Reboot to activate the kernel modules.
2. Configure Docker & toolkit
Install Docker Engine then add the NVIDIA Container Toolkit: `sudo apt install -y docker.io nvidia-container-toolkit` followed by `sudo nvidia-ctk runtime configure --runtime=docker`.
3. Enable persistence
Run `sudo nvidia-smi -pm 1` and confirm `nvidia-smi` reports each GPU. Optional: automate via systemd so GPUs stay initialised after restarts.
4. Deploy the agent
From the Provider Portal download the install script or run the curl command shown above with your API key. Each rig registers with its Mafutai zone.

Windows hosts

1. Install Windows GPU drivers
Download the latest Studio/Data Center drivers from NVIDIA (or AMD ROCm for Windows if supported). Reboot after installation.
2. Enable WSL2
Open PowerShell as Administrator and run `wsl --install`. Choose Ubuntu 22.04 LTS when prompted. This provides the Linux environment required by the provider agent.
3. Install Docker Desktop
Enable the WSL2 backend and GPU support in Docker Desktop settings. Ensure the Distribution (Ubuntu) is checked under Resources → WSL Integration.
4. Configure CUDA inside WSL
Inside Ubuntu run `sudo apt update && sudo apt install -y nvidia-cuda-toolkit` and verify `nvidia-smi` works. Follow the NVIDIA guide for CUDA on WSL if additional packages are needed.
5. Run the Mafutai agent
From within WSL execute the provider agent install script using your API key. The agent runs as a background service and reports heartbeats to Mafutai.

Tip: For Windows, keep the provider agent inside WSL2 where Linux tooling is available. Production providers generally graduate to dedicated Linux hosts for lower latency and easier automation.

Need a hand?

Reach out to the Mafutai ops team for architecture reviews, pricing guidance, or bulk onboarding support.

support@mafutai.co.za Join the early access list