Install with npm. Paste a handoff token. You're SSH'd into a real broken server from your own iTerm, Warp, or whatever terminal you use daily. Your keybindings, your split panes, a real outage to debug.
$ parium open █▀█ ▄▀█ █▀█ █ █ █ █▀▄▀█ █▀▀ █▀█ █▀▄ █ █▄█ █ ▀ █ Chaos Terminal Client v0.1.0-alpha.7 Paste handoff token: •••••••••••• ✓ Token validated ✓ Session resolved - k8s-chaos-war-room ⟳ Attaching to terminal... ────────────────────────────────────── SESSION K8s Cascading Failure STATUS ● LIVE PHASE 3 of 6 - DNS network policy IMPACT $120K/hr ────────────────────────────────────── candidate@prod-worker-07:~$ █
Senior engineers have workflows built over years. Shell aliases, split panes, muscle memory. The CLI lets them work in the environment they already know, so you're measuring debugging ability, not how well they adapt to an unfamiliar interface.
The flow is deliberately minimal. Install once, launch with parium open, paste a token, debug.
One npm command. Native terminal access on the operating systems your team already uses, with a lightweight Node runtime requirement.
The browser lobby generates a secure token when a candidate or engineer is ready. Paste it into the CLI and the session resolves automatically.
WebSocket attaches to the live container. Same environment as the browser flow, delivered to your native shell.
Lobby UI, reconnect logic, presence tracking. More than a wrapper.
WebSocket connection to real containers. Raw PTY I/O, so your terminal renders the output instead of a browser emulator. Resize and signal forwarding work as expected.
The browser lobby generates time-limited tokens. The CLI validates them server-side and resolves the correct session, scenario, and room automatically.
For Chaos Mode sessions, the CLI renders an interactive lobby with presence indicators, readiness gates, and hotkey controls. Press S to start, Tab to cycle themes.
Network drops happen. The CLI retries with exponential backoff (up to 5 attempts), re-validates the session, and picks up where you left off.
Dark, light, and mono themes. Auto-detects weak terminals (no Unicode, no color) and falls back to mono. Override with --theme or Tab in the lobby.
Keep the browser dashboard open for telemetry and room context while you work in the terminal. Both stay in sync.
The CLI preflight checks your terminal capabilities, pings the server, and reports latency before you even paste the token. When the session is ready, it attaches and your prompt appears.
$ npm install -g @parium.ai/cli@latest
--json
PARIUM / preflight ✓ theme dark ✓ terminal iTerm2 · truecolor · 120×32 ✓ server https://parium.ai · 42ms PARIUM / session ✓ Token validated ✓ Room resolved - chaos-room-42 ✓ Terminal attached ╭──────────────────────────────────────╮ │ SESSION Azure VM 503 │ │ STATUS ● LIVE │ │ ROOM Engineer A + Engineer B │ ╰──────────────────────────────────────╯ root@edge-api-vm:~$ █
No. The CLI is another way to access the same incident environments. Candidates or engineers can choose between the browser terminal or the CLI. Both connect to the same live container running the same scenario.
When realism matters. Senior infrastructure candidates have years of terminal muscle memory. The CLI removes the browser abstraction so you see how they actually debug, in their own environment with their own tools. It's especially good for Chaos Mode war rooms where engineers want split-pane terminal control alongside browser telemetry.
The browser lobby generates a time-limited, single-use token when the engineer is ready. They copy it and paste it into the CLI. The CLI validates the token server-side and resolves the correct session: scenario, room, engineer role. The token auto-detects from clipboard if available.
Yes. The CLI attaches to the same containers as the browser flow. GPU failures, Kubernetes cascading failures, Azure networking, Docker security - they all work identically. The incident engine is the same regardless of how you connect.
The CLI retries automatically with exponential backoff, up to 5 attempts. It re-validates the session and reconnects to the container. The candidate's work is preserved, so they pick up where they left off.
The CLI says you understand how infrastructure engineers actually work.