● v0.1 · Apache 2.0 · runs on your GPU

Your GPU. Your team.
No cloud tax.

Apex is a self-hosted ML platform for small AI teams. One pip install, full job queue, browser-native VS Code, real-time GPU monitoring — all running on the workstation you already own.

$ pip install apex && apex start click to copy
View on GitHub → See the features
Python 3.10+ · Linux · NVIDIA GPU (optional)
The math

Stop paying rent
on hardware you could own.

Compare a full month of training compute, 24/7.

RunPod RTX 4090 $0.44/hr × 720 hrs
$316 / mo
Lambda H100 $2.99/hr × 720 hrs
$2,150 / mo
AWS p4d.24xlarge (8× A100) $32.77/hr × 720 hrs
$23,594 / mo
Your workstation + Apex Team tier, 8 seats, unlimited jobs
$29 / mo
Features

Everything a small ML team
actually needs.

No Kubernetes, no cloud bill, no DevOps engineer required. Just the things that matter.

Apex overview dashboard

Live overview dashboard

Jobs today, GPU hours used, queue depth, success rate — at a glance. Updates in real-time via server-sent events.

Live training logs streaming from a running container

Live training logs

Click any running job to attach a WebSocket log stream. See loss curves, step counts, and OOM warnings as they happen.

Dev sessions list

Browser VS Code in one click

Launch a code-server container on a free port with your workspace pre-mounted. Full VS Code, no install, GPU attached.

GPU metrics dashboard

Real GPU telemetry

pynvml under the hood. GPU util, VRAM used/total, temperature, power draw, CPU, RAM — every 2 seconds, in the topbar and on the metrics page.

Job history table

Full job history

Sortable table of every job you've ever run. Filter by status, tail logs for any of them, cancel running jobs, remove old ones.

Docker images list

Native Docker integration

Reads directly from your Docker daemon — no registry push required. Pre-built apex/code-server images for Python and PyTorch-CUDA included.

Three steps

From install to training
in under 60 seconds.

No YAML. No helm charts. No IAM roles.

01

Install

pip install apex — registers the apex CLI. Python 3.10+ and a running Docker daemon are the only prerequisites.

02

Start

apex start — boots the platform on port 7000, starts the GPU monitor, opens your browser automatically.

03

Submit a job

Paste a Docker image, an entry script, click Queue job →. Apex pulls, runs, streams logs, and reports success — all on your GPU.

The stack

Intentionally boring.

Python 3.10+ · pynvml · psutil · vanilla HTML/CSS/JS

React · Redis · Postgres · RabbitMQ · Kubernetes · Helm · Node.js · webpack · Terraform

One Python process. One pip install. One GPU machine.

Pricing

Free forever.
Paid when you have teammates.

All tiers run on your hardware. You're not renting compute, you're not renting seats — you're renting the multi-user bits.

Free
$0
forever · self-hosted
  • Full feature set
  • 1 seat
  • Unlimited jobs
  • Unlimited dev sessions
  • All Docker images
  • Community support (Discord + GitHub)
Install with pip
Team
$29/ mo
flat · up to 8 seats
  • Everything in Free
  • Up to 8 seats
  • Multi-user auth (JWT + SSO)
  • Audit log
  • Per-user dev sessions
  • Priority email + Discord support
  • License key
Start a 14-day trial →
Hosted
$99/ mo
starting · + GPU pass-through
  • Everything in Team
  • We host it on a rented GPU
  • No install, no maintenance
  • SLA
  • Daily backups
  • Custom domain
Contact us