One command installs everything — Docker, NVIDIA Container Toolkit, and the agent. Takes 5 minutes on any Linux machine or Windows PC.
5 min
Typical install time
1
Command to install
0
Config files to edit
24/7
Auto-restart on crash
✅ Signed in — your commands below are personalised
They carry a single-use install token, so the installer will not ask you anything.
Your machine registers to your account automatically.
Create your account first — it takes about 30 seconds
You'll get a personalised one-line command. Paste it once and your machine
is registered and earning — no details typed into a terminal, and you can sign in to
your dashboard to see what you've made.
Tested on Ubuntu 20.04, 22.04, 24.04. Installs Docker, NVIDIA Container Toolkit, and the node agent automatically.
Requirements
NVIDIA GPU (RTX/GTX/Quadro) — CUDA required
Ubuntu 20.04+ or Debian 11+
NVIDIA drivers 525+ (installer will check)
50+ Mbps internet connection
Sudo access
AMD ROCm GPU? Use the
Step 1 — Get your personalised command
Your command carries a single-use setup token, so the machine registers
to your account automatically and asks you nothing.
A generic command cannot work — it has no way to know
whose account to pay.
The installer auto-detects your AMD GPU, installs ROCm 6.1, installs PyTorch with ROCm support, registers your node, and sets up the systemd service. Takes 10–15 minutes.
Native macOS agent (no Docker). Works on Mac Studio, Mac Pro, MacBook Pro. Installs PyTorch MPS, Apple MLX, HuggingFace Transformers, and the node agent — all in one command. No sudo required.
Requirements
Apple Silicon chip: M1, M2, M3, or M4 (any variant — base/Pro/Max/Ultra)
macOS 12 Monterey or newer (macOS 13+ recommended for MLX)
Python 3.10+ (installer will offer to install via Homebrew if missing)
50+ Mbps internet · Mac plugged in during rentals
Intel Macs supported at lower tier (CPU workloads only)
Detects your chip (M1→M4, Pro/Max/Ultra), installs PyTorch MPS + MLX + HuggingFace, registers your node, and installs a launchd service that starts on login. Takes 5–15 minutes depending on internet speed.
Step 2 — Verify the service is running
launchctl list | grep gridshare
Step 3 — Check logs
tail -f ~/.gridshare/node.log
💡 Apple Silicon Pricing (Live Beta)
Pricing by chip — ranked by memory bandwidth × unified memory:
• M3 / M2 / M1 Ultra → ₹44–₹53/hr (M3 Ultra tops the ladder — you earn ₹37.10/hr)
• M5 Max / M4 Max / M3 Max / M2 Max → ₹35–₹40/hr
• M5 Pro / M4 Pro / M1 Max → ₹31–₹33/hr
• M3 Pro / M2 Pro / M1 Pro → ₹22–₹26/hr
• M5 / M4 / M3 / M2 / M1 base → ₹12–₹17/hr
Killer use case: M3 Ultra with up to 512 GB unified memory runs 70B LLMs that no consumer NVIDIA GPU can hold. Service starts on Mac login · auto-pauses when you use Mac · resumes when idle. You earn 70% of every rate above.
Double-click it, then click Yes on the Windows prompt. Windows restarts once to enable WSL2, then GridShare finishes by itself — no commands. (If SmartScreen warns: More info → Run anyway.)
Requirements
Windows 10 (version 2004+) or Windows 11
NVIDIA GPU with drivers 525+
WSL2 enabled (installer does this automatically)
PowerShell 5.1+ (pre-installed on Windows 10+)
Must run PowerShell as Administrator
Step 1 — Open PowerShell as Administrator, run:
irm https://gridshare.in/dl/install.ps1 | iex
The installer will:
Enable WSL2 if not already active (restart may be required)
Install Ubuntu 22.04 in WSL2
Install Docker + the NVIDIA Container Toolkit inside WSL2
The GridShare node agent is a native process (not a container) — it needs host-level access to manage buyer workload containers, so it can't run containerized itself. If Docker and the NVIDIA Container Toolkit are already installed, the installer detects and reuses them.
Prerequisites (installer checks these, installs if missing)
Docker Engine 24+ (already installed — installer will use it)
NVIDIA Container Toolkit (nvidia-docker2)
NVIDIA drivers 525+
Run the installer
Same as every other platform: grab your personalised one-line command
from the portal and paste it. The installer detects your existing Docker
and NVIDIA toolkit and reuses them.