Install these before the first launch:

• Docker with Docker Compose v2.20.3 or newer (v2.26+ recommended). The top-level docker-compose.yml merges the per-service fragments with Compose's include: directive, which older releases do not support. ./start.sh checks the version and stops if it is too old.
• uv, or Python 3.10 or newer. ./start.sh prefers uv run and falls back to the system python3.
• Git.

Clone the repository and run every command from its root (atlas/):

git clone https://github.com/thekaveh/atlas && cd atlas
./start.sh

The setup wizard asks for track and profile, base port and project name, service SOURCE choices and host aliases. It then shows the launch summary.

./start.sh
./start.sh --track gen-ai-rag
./start.sh --track data-eng
./start.sh --base-port 64000
./start.sh --setup-hosts

Flags such as --base-port, --setup-hosts, --detach or any --*-source skip the wizard. Atlas then launches what .env holds (the .env.example defaults on a fresh clone), not a track's subset. Add --track <key> to apply one.

After the launch, open the Atlas root dashboard at http://localhost:<BASE_PORT>. The default is http://localhost:63000; after ./start.sh --base-port 64000 it is http://localhost:64000. KONG_HTTP_PORT in .env holds the dashboard port, which is derived from BASE_PORT. Direct URLs and Kong aliases are in the service catalog and the ports and routes reference.

There is no single login. Most dashboards open with an administrator password that Atlas writes to .env on first start:

• Open WebUI: OPEN_WEB_UI_ADMIN_EMAIL / OPEN_WEB_UI_ADMIN_PASSWORD.
• LiteLLM: admin / LITELLM_MASTER_KEY.
• Grafana: admin / GRAFANA_ADMIN_PASSWORD.
• Supabase Studio and several other dashboards: the Kong pair DASHBOARD_USERNAME / DASHBOARD_PASSWORD.

n8n asks you to create its owner account. With an empty JUPYTERHUB_TOKEN (the default), JupyterHub generates a token and prints it in its container log. Access and Credentials lists every surface and the credential that opens it.