# Cua Fleet (Windows Server)

Official pricing: https://cua.ai/pricing  
Category: windows · Isolation: vm

## Pricing regimes (raw)

- **Cua Fleet usage-based, Windows Server 2022 image** (resource): $0.044625/vCPU-h, $0.0223125/GiB-h
- **Cua Cloud macOS (separate product)** (sizes): Cloud macOS ? vCPU/? GiB $null/h [sales]

## Features

Yes: ≥24 h sessions, custom image (Docker or snapshot), full VM (own kernel), Docker inside, browser, desktop GUI, computer-use API, browser + desktop control, preinstalled agents, HTTPS preview URLs, self-hosting, BYOC, open source, SOC 2, Windows, Python, Node.js, MCP server, gVisor or VM (no shared kernel), inbound access rules, firewall inside (nftables)

No: snapshots, memory snapshots, fork/clone, pause/resume, persistent disk, volumes, idle auto-stop, start from your own snapshot, your own Docker/OCI image, systemd, code interpreter, anti-bot stealth, CAPTCHA solving, residential IPs, SSH, public IPv4, custom domains, raw TCP inbound, egress allowlist, static egress IP, private networking, HIPAA, SSO, GPU, arm64, macOS, GPU desktop, HTTP method/path egress rules, wake on request, live resize, live fork (no pause), memory fork, EU data residency, automatic snapshots, snapshots on demand, agent harness API, hosted agent API (their own agent), secret proxy, secret proxy for any API, extra volumes, shared volumes

Unknown: everything else. Evidence (source + quote) per feature: https://battleships.dev/data/providers/cua-windows.json → feature_evidence

## Caveats

- Billing granularity, minimums, storage, egress and paused-state billing are not published.
- No Windows surcharge published; if one exists the estimate is low.
- macOS is NOT hosted by Cua (local-only via Lume on your own Mac).
- A 'Pro from $10/mo' plan appears only in third-party guides; not encoded.
- Releasing a claim does not delete or scale down its pool: retained warm sandboxes can keep billing until the pool is scaled down, deleted or expires.

## How this provider charges


Cua Fleet is Cua's hosted cloud for computer-use sandboxes; the Cua SDK itself is open source (MIT). It bills one usage rate per vCPU-hour and per GiB-hour of memory for every OS. The Fleet's built-in images are **Ubuntu 24.04 and Windows Server 2022**. Windows 11, macOS and Android are rejected for Fleet; they are local-only through Lume, QEMU or an emulator.

## Regime table

| # | Regime | When it applies | How billed | Numbers | Source |
|---|---|---|---|---|---|
| 1 | Fleet usage, Windows Server 2022 | Cloud sandbox | Allocated vCPU + memory per hour. Granularity/minimum not published | **$0.044625/vCPU-h + $0.0223125/GiB-h**. No published Windows surcharge. 4 vCPU/8 GiB = **$0.357/h** | https://cua.ai/pricing, https://cua.ai/docs/reference/sandbox-sdk/runtime-support |
| 2 | Local sandbox (OSS) | Your machine | Free (Docker, QEMU, Apple VZ; Windows 11 via local ISO; macOS via Lume) | $0 | https://cua.ai/pricing |
| 3 | Enterprise / on-prem / BYOC | Contact | custom | custom | https://cua.ai/pricing |

## Gotchas

1. Storage, egress, idle/paused billing, minimums and free credits are **not published**.
2. No published Windows licence surcharge. If one exists, this estimate is low.
3. Cua does **not** host macOS. The sweep's "macOS only locally through Lume" holds.

## Worked example

Workload: 4 vCPU / 8 GiB, 50 concurrent × 8 h/day × 22 days (8,800 h), 30% CPU, 50 GiB snapshots, 100 GiB egress.

- Compute: 8,800 × $0.357 = **$3,141.60/month**.
- Snapshots, egress and stopped-state billing are unpublished (null), so the real total is at least $3,141.60.

Sources: https://cua.ai/pricing · https://cua.ai/docs/reference/sandbox-sdk/runtime-support