$CLI$CLI
// coming soon · Q3

Deploy agents
on rented metal.

Push Claude, OpenClaw, vLLM, or any Docker image onto a GPU you've rented through $CLI. One click. Encrypted secrets. Streaming logs. USDC- and SOL-billed by the second.

deployingrig-04 · RTX 4090
  • ▸ pull imageclaude-code:latestok
  • ▸ mount secrets8 keysok
  • ▸ expose ports22, 8000, 8888ok
  • ▸ warmup gpu1× RTX 4090running
  • ▸ stream logsws://…pending
$ cli deploy --runtime claude-code --rental rnt_8x2y

// // supported runtimes

Drop-in agents for any workload

Each runtime ships with a verified base image, port map, and secret schema. Bring your own image when ours don't fit.

Claude CodeAnthropic

Headless coding agent in your pod

  • ▸ Tool use
  • ▸ Shell + file access
  • ▸ Persistent context
queued · q3
OpenClawOpen

Open-source autonomous agent

  • ▸ Web browse
  • ▸ Self-hosted
  • ▸ Plugin system
queued · q3
vLLM ServerInference

OpenAI-compatible inference

  • ▸ Llama · Qwen · Mistral
  • ▸ Auto-batching
  • ▸ OpenAI API
queued · q3
Custom DockerBYO

Bring your own image

  • ▸ Any image, any port
  • ▸ Secrets injection
  • ▸ Health checks
queued · q3

// // workflow

From idle GPU to live agent in five steps

  1. 01Pick a rental

    Select any active GPU you've rented through $CLI.

  2. 02Pick a runtime

    Claude Code, OpenClaw, vLLM, or upload a Dockerfile.

  3. 03Inject secrets

    Encrypted env vars, scoped to the pod. Rotate anytime.

  4. 04Launch

    $CLI provisions, exposes ports, opens a web console.

  5. 05Monitor

    Logs, GPU util, cost-per-hour, auto-stop on cost cap.

// // preview

Operate like ops, not ClickOps

Same primitives you already know — pods, logs, secrets, ports — wrapped in a one-line CLI and a live web console.

cli · rig-04
$ cli rentals ls
  rnt_8x2y · RTX 4090 · active · 1.42h · 0.84 USDC
$ cli deploy --runtime claude-code --rental rnt_8x2y
  ↳ pulling anthropic/claude-code:latest …
  ↳ writing /etc/cli/secrets (8 keys) …
  ↳ exposing :22 :8000 :8888 …
  ↳ ready ssh://root@1.2.3.4:22022
$ cli logs rnt_8x2y --follow
  [12:04:01] agent: starting plan…
  [12:04:02] tool: read_file src/router.tsx (2.4kB)
  [12:04:04] tool: edit_file src/router.tsx +12 -3
  [12:04:07] agent: ok, tests pass.

Connection details

SSH

ssh root@1.2.3.4 -p 22022

Jupyter

https://abc.cli.dev/?token=…

App

https://abc-8000.cli.dev

Logs

wss://logs.cli.dev/rnt_8x2y
Connection details only appear after successful provisioning. Failed pods refund fully and never charge for unused time.

// // roadmap

What's queued behind launch

▸Prompt presets
▸Encrypted secret vault
▸Custom port exposure
▸Live log streaming
▸Auto-restart on crash
▸Cost-cap kill switch
▸Multi-replica fleets
▸Webhook triggers
▸Cron schedules
// waitlist open

Be first to deploy.

We'll email when Deploy Agent goes live. Early signups get a credit grant and priority access to managed runtimes.