# Orgo

Official pricing: https://www.orgo.ai/pricing  
Category: agent-sandbox · Isolation: vm

## Pricing regimes (raw)

- **Monthly plan pool (Linux)** (pool)
- **Monthly plan pool (Windows, Scale plan only)** (pool)
- **Yearly billing (Linux)** (pool) [sales]
- **Yearly billing (Windows, Scale only)** (pool) [sales]
- **Enterprise (Linux + Windows + Mac, 60+ computers)** (pool) [sales, beta]
- **Enterprise GPU/hardware fleet** (pool) [sales]

## Features

Yes: snapshots, memory snapshots, fork/clone, pause/resume, persistent disk, idle auto-stop, ≥24 h sessions, custom image (Docker or snapshot), start from your own snapshot, full VM (own kernel), Docker inside, root, browser, desktop GUI, computer-use API, browser automation API, desktop control API, browser + desktop control, residential IPs, preinstalled agents, SSH, HTTPS preview URLs, egress allowlist, open internet, static egress IP, self-hosting, BYOC, GPU, Windows, macOS, Python, Node.js, GPU desktop, credential injection, wake on request, live resize, live fork (no pause), memory fork, webhooks, scoped API keys, spend limits, MCP server, snapshots on demand, hosted agent API (their own agent), gVisor or VM (no shared kernel), inbound access rules, firewall inside (nftables)

No: volumes, your own Docker/OCI image, anti-bot stealth, CAPTCHA solving, public IPv4, open source, arm64, HTTP method/path egress rules, EU data residency, automatic snapshots, agent harness API, secret proxy, secret proxy for any API, extra volumes, shared volumes

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

## Caveats

- Hacker tier vCPU is not published (live pricing page 2026-09-28: "1 computer, 8GB RAM. 40GB storage", no CPU), so its pool vCPU is null and it is not auto-picked. Per the instance-types doc an 8 GB computer is the Medium 2 CPU / 8 GB shape; 4 CPU comes with 16 GB+. Startup/Scale vCPU remain the computers x 4 upper bound. Per-computer disk within the storage pool is not published.
- Pool vCPU per tier not published; set to computers x 4 vCPU (per-computer hard cap), so large-vCPU workloads may be accepted that the plan would refuse (403 quota).
- RAM pool per plan: 8 / 32 / 128 GB. A 4 vCPU / 8 GB workload fits 1 / 4 / 16 computers.
- Workloads above 16 concurrent computers need Enterprise (custom price); stacking several Scale plans is not documented.
- Windows only on Scale ($399/mo) or Enterprise; macOS only in Enterprise closed beta. Windows licence not itemised.
- Stopping a computer archives its disk (new host + new IP on start, no memory state). Idle time is not billed separately because billing is flat.
- Capacity add-ons ("add computers as demand grows"), egress, IPv4 not priced publicly (null).
- Monthly price verified on live page; yearly prices read after toggling "Yearly": $24/$82/$332 per month, "$290 / $990 / $3,990 billed yearly".
- Category changed windows -> agent-sandbox (Linux is the default OS on all paid tiers; Windows only on Scale). Yearly modes flagged sales (annual prepay, opt-in). Retained computer slots count ALL existing computers (stopped ones too) until deleted, not only running peak; stopping does not reduce the fee. Per-computer CPU values accepted: 0.5/1/2/4; RAM 4-64 GB; default 1 CPU/4 GB. (patch v3-computer-use, 2026-09-28)
- Capacity add-ons (official changelog 2026-07-13, not re-rendered live): RAM $4/GB, vCPU $5 per half-core, disk $2 per 20 GB, extra computer slot $1 (up to 30); billing period not stated, so not modelled as pool tiers. Pre-July subscribers keep their old plan/limits. (patch v8-history, 2026-09-28)
- Fleets above 16 computers (e.g. 50 x 4/8) need Enterprise (60+ computers, custom price) or capacity add-ons whose billing period is unstated; not priceable from public data.

## How this provider charges


Orgo sells persistent "computers for agents": full desktops with screenshot/click/type/scroll/drag APIs, REST/SDK/CLI/MCP and SSH. Pricing is a **flat monthly plan** that buys a number of always-on computers sharing a RAM and storage pool. There is no per-hour meter.
- **OS:** Linux on every plan; Windows from the Scale plan; Mac desktops only in Enterprise closed beta.
- **Sizes:** a computer is sized by `cpu` (0.5, 1, 2 or 4 cores) and `ram` (4, 8, 16, 32 or 64 GB). The hard cap per computer is 4 vCPU / 64 GB / 300 GB, and each plan sets a lower per-computer ceiling (unpublished).

## Regime table

| # | Regime | When it applies | How billed | Numbers | Source |
|---|---|---|---|---|---|
| 1 | Hacker | 1 computer, Linux | Flat monthly | **$29/mo**: 1 computer, 8 GB RAM, 40 GB storage, 1 seat, $5 AI credit | https://www.orgo.ai/pricing |
| 2 | Startup | Up to 4 computers, Linux | Flat monthly | **$99/mo**: 4 computers, 32 GB RAM, 160 GB, 5 seats, $10 AI credit | same |
| 3 | Scale | Up to 16 computers, **Linux + Windows** | Flat monthly | **$399/mo**: 16 computers, 128 GB RAM, 640 GB, 25 seats, $50 AI credit, custom templates | same |
| 4 | Yearly billing | Toggle "Yearly" | Prepaid annually | **$290/yr** ($24/mo), **$990/yr** ($82), **$3,990/yr** ($332) | same (Yearly toggle) |
| 5 | Enterprise | 60+ computers, Linux + Windows + **Mac** | "Custom. Volume pricing, billed annually." Shared, dedicated or on-prem | null | same |
| 6 | Stopped computer | `stop` | Disk archived. Restarts on a fresh host with a new IP and no RAM state | Included in the plan | https://docs.orgo.ai/guides/instance-types |
| 7 | Capacity add-ons | "Add computers as demand grows" | Not priced | null | pricing |
| 8 | AI credit | Managed model access | Model-usage credit, **not compute** | $5 / $10 / $50 per month | pricing |
| 9 | Egress / IPv4 | – | Not published | null | – |

## Gotchas

1. **The plan caps both count and RAM.** Scale's 128 GB pool means 16 × 8 GB computers, or only 4 × 32 GB.
2. **Pool vCPU is unpublished.** The card uses computers × 4 vCPU (the hard per-computer cap) as an upper bound. Real plan ceilings are lower and can return 403 `PER_COMPUTER_CPU_CAP`.
3. **Windows needs Scale ($399).** No Windows licence line is shown.
4. **macOS is closed beta on Enterprise only**, with no price.
5. **Flat pricing:** idle time costs nothing extra, and a busy fleet costs no more than an idle one.
6. **Stacking plans is undocumented.** More than 16 concurrent computers officially means Enterprise.

## Worked example

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

| Regime | Fits? | Monthly |
|---|---|---|
| Scale monthly | 16 computers only (8 GB each uses the whole 128 GB pool) | $399 covers 16 concurrent |
| 4 × Scale (hypothetical; stacking undocumented) | 64 computers / 512 GB | 4 × $399 = **$1,596** (yearly: 4 × $332.50 = $1,330) |
| Enterprise (the official path for 50) | yes | null (custom) |
| Windows | Scale or Enterprise only | same as above |
| Snapshots / egress | not priced | null ($0 assumed) |

Hours and CPU utilisation do not change the bill.

Sources: https://www.orgo.ai/pricing · https://docs.orgo.ai/guides/instance-types · https://docs.orgo.ai/llms.txt