# Orgo Pricing

Pricing definitions and annual billing display are maintained alongside the
public plan UI. The prices below are the current published plan rates.

## At a glance

| What                                      | Price                                  |
| ----------------------------------------- | -------------------------------------- |
| Small Linux (4 GB, 0.5 vCPU), by the hour | $0.023/hr running, $0.0014/hr stopped  |
| Medium Linux (8 GB, 2 vCPU), by the hour  | $0.055/hr running, $0.0027/hr stopped  |
| Large Linux (16 GB, 4 vCPU), by the hour  | $0.11/hr running, $0.0055/hr stopped   |
| XL Linux (32 GB, 6 vCPU), by the hour     | $0.202/hr running, $0.011/hr stopped   |
| Starter plan                              | $29/mo or $276/yr                      |
| Growth plan                               | $99/mo or $948/yr                      |
| Scale plan                                | $399/mo or $3,828/yr                   |
| Residential IP data past what a plan includes | $3/GB from credits                 |
| Public URLs for apps on a computer        | Included, 10 a computer (see [Hosting](#hosting)) |

A new account can get $3 of starting credit (see [Free Tier](#free-tier) for who gets it). Hourly is billed by the second from credit; a free account's hourly computer is a Medium; other sizes need a paid plan. 2XL and Enterprise are on request. Details below.

Provision and manage AI agent computers (virtual desktops) in the cloud. Connect via VNC, integrate with AI agents, and scale from prototype to production.

**Free to sign up.** Anyone can create a free Orgo account, explore the dashboard, create workspaces, and generate API keys at no cost. A new account that passes the free checks gets $3 of starting credit, which runs one Medium Linux computer by the hour (see [Free Tier](#free-tier)). Other sizes need a paid plan. More computers need a paid plan or credit you add: then every computer counts against the account's 50, and past 50 more run by the hour from credit up to 999 (a free account's are Medium Linux). Free users invited into a paid workspace can create and use computers there, funded by that workspace's plan, no separate subscription required.

## Plans

Plans are described by the computers they keep always on. A Medium is 2 vCPU,
8 GB RAM and 40 GB disk.

| Plan       | Monthly | Yearly    | Runs always on                         | Teammates besides you | For                                                    |
| ---------- | ------- | --------- | -------------------------------------- | ------------------ | -------------------------------- |
| Pay as you go | $0   | $0        | None: computers by the hour from credit | 5                 | No plan: like a model API's usage billing |
| Starter    | $29/mo  | $276/yr   | 1 Medium computer                      | 5                  | For individuals |
| Growth     | $99/mo  | $948/yr   | 4 Medium computers, or mix sizes up to XL    | 5            | For agencies |
| Scale      | $399/mo | $3,828/yr | 16 Medium computers, or mix sizes up to XL   | 25           | For agent apps |
| Enterprise | Custom  | Custom    | Custom                                 | No limit           | Large fleets, custom terms, SLA  |

Teammates are the people who work in your workspaces. The people your agents
or your app serve (clients, customers, end users) never need a seat, on any plan.

Need more? Add a Medium to any plan for $32 a month, or run one by the hour
from $0.055 (see [By the hour](#by-the-hour)). Up to 50 computers on a plan,
999 with credit.

The $29 Starter plan is one Medium computer that stays on all month. The same
computer by the hour costs about $40 a month ($0.055 x 730 hours). When your
hourly computers cost more than a plan that holds them, Settings › Plan &
billing says so, with that plan one press away; once you have it, they move
onto it.

Plans bill monthly by default, and a first plan is monthly. Once you have a
plan, switching it to yearly saves 20% on every plan, shown in the app as the
per-month equivalent ($23, $79 and $319) with the annual total beside it.
Add-ons bought yearly save the same 20%.

Changing plans keeps your one subscription. An upgrade, or a switch from
monthly to yearly, applies now: the unused part of the current period is
credited and the difference is charged today. A downgrade, or a switch from
yearly to monthly, starts at the end of the period. Cancelling also takes
effect at the end of the period; until then you can keep the plan or switch
to a smaller one instead (Settings › Plan & billing). A monthly plan can pause
for 1 to 3 months instead: from the end of the period nothing is charged and
the account is on Free, then the plan comes back on its own (or sooner with
Resume now). A plan with a dedicated IP or a custom invoice can't pause.

Paid subscriptions start at checkout. Some referral offers include a subscription trial; any such trial requires a card, and its duration and subsequent charge are shown at checkout.

Business buyers can enter a VAT or other tax ID at checkout. It is saved to their account and shown on their invoices.

The computer trial is no longer offered to new accounts: a free account runs its $3 starting credit on an hourly computer instead (see [Free Tier](#free-tier)). This is separate from any referral subscription trial.

## Computers

Virtual desktops that run 24/7, accessible via browser or API.

Each paid plan keeps a number of computers always on: **Starter 1 Medium,
Growth 4 Mediums, Scale 16 Mediums**, or fewer bigger ones (a Large takes the room of two Mediums, an XL of
four). Underneath, a plan includes a vCPU and
RAM allowance that all of the account's computers share;
add more of either at any time (see [Add-ons](#add-ons-pay-as-you-grow)). Past 50, more
computers run by the hour from your credit (see [By the hour](#by-the-hour)),
up to 999 an account. Every computer counts, running or stopped, and there is
no limit on how many run at once. Free runs hourly computers only (see
[Free Tier](#free-tier)).

| Tier       | Mediums always on | vCPU included | RAM included (all computers) | Biggest computer | Computers (count) |
| ---------- | ----------------- | ------------- | ---------------------------- | ---------------- | ----------------- |
| Starter    | 1                 | 2             | 8 GB                         | Medium (8 GB)    | 50, 999 with credit |
| Growth     | 4                 | 8             | 32 GB                        | XL (32 GB)       | 50, 999 with credit |
| Scale      | 16                | 36            | 128 GB                       | XL (32 GB)       | 50, 999 with credit |
| Enterprise | Custom            | Custom        | Custom                       | Custom           | Up to 999         |

The count (50) is how many computers an account may have, not how many a plan
runs: past the plan's allowance a computer needs an add-on or runs by the hour.

Legacy plans that are no longer sold keep their own vCPU and RAM and the same
count of 50.

All paid plans use always-on persistent computers. Computers are never auto-suspended. They stay running until you stop or delete them.

## What more computers cost

A computer draws on the plan's vCPU and RAM. Past them, the rest is bought as
add-ons at the component rates below (a Medium's worth, 8 GB and 2 vCPU, is
$30; with its disk, $32), or the computer runs by the hour from your credit
(see [By the hour](#by-the-hour)). There is no per-computer
fee. The cheapest setup for a few counts of Medium Linux computers:

| Computers | Cheapest setup      | Monthly   | Yearly, per month |
| --------- | ------------------- | --------- | ----------------- |
| 1         | Starter             | $29       | $23               |
| 5         | Growth + add-ons    | $129      | $103              |
| 10        | Growth + add-ons    | $279      | $223              |
| 25        | Scale + add-ons     | $649      | $519              |
| 50        | Scale + add-ons     | $1,399    | $1,119            |

From 14 computers Scale costs no more than Growth plus add-ons, and it includes
more (Windows and 25 team members). Every plan can own up to 9999 workspaces. Past 50 computers, each new
one runs by the hour from credit; Enterprise prices by volume.

### By the hour

A Linux computer past the plan's vCPU and RAM can also run by the hour from
your credit instead (your whole balance: credit you bought, starting credit and
your plan's monthly platform credits), with no add-on to buy; a Free
account's computer always does (see
[Free Tier](#free-tier)). So does every computer past an account's 50: a new
one is made as long as the credit covers at least an hour of all the account's
hourly computers, the new one included, up to 999 computers an account. Running is the plan's rate plus 25%: the same RAM,
vCPU and disk rates, times 1.25, over 730 hours a month, rounded to the
nearest $0.001. Stopped is the disk alone at the plan's rate, with no markup:
$1 per 20 GB a month over 730 hours, rounded to the nearest $0.0001. Billed by
the second, with no hourly minimum:

| Size   | RAM   | vCPU | Disk   | Running   | Stopped    |
| ------ | ----- | ---- | ------ | --------- | ---------- |
| Small  | 4 GB  | 0.5  | 20 GB  | $0.023/hr | $0.0014/hr |
| Medium | 8 GB  | 2    | 40 GB  | $0.055/hr | $0.0027/hr |
| Large  | 16 GB | 4    | 80 GB  | $0.11/hr  | $0.0055/hr |
| XL     | 32 GB | 6    | 150 GB | $0.202/hr | $0.011/hr  |
| 2XL    | 64 GB | 8    | 300 GB | $0.368/hr | $0.0205/hr |

Sizes are offered as for the plan: Small through the API only, 2XL on request.

**Changing size.** A running hourly Linux computer on a paid plan, or on an account that pays from credits, can move to any size above, up or down, and pays the new rate from that moment. Its disk stays as it is, so running is the size's rate on that disk and never under this table, and stopped is that disk. A higher rate needs credit for an hour of it and an hour of your other hourly computers. A computer with an end stays prepaid to that end: a bigger size takes the extra from your credit, and a smaller one returns the unused part. A Free account's computer stays Medium, and Windows and GPU computers keep their size. API: `GET` and `PATCH /api/computers/{id}/resize`.

## Compute Resources (per computer)

Every computer starts with the same baseline; add more RAM, vCPU, or disk to your account as an add-on.

The app's create form starts at a **Medium: 8 GB RAM, 2 vCPU, 40 GB storage**,
on every plan. Through the API (`POST /api/computers`), a create that names no
size gets **4 GB RAM and 1 vCPU**; one that names only `ram` gets that size's
vCPU (8 GB: 2, 16 GB: 4, 32 GB: 6, 64 GB: 8); an explicit `cpu` is used as
given. How far one computer can grow without an add-on depends on the plan. Its RAM
can be as much as the plan's whole RAM total, up to XL on your own; storage has
a per-computer maximum:

| Tier       | RAM (max)                      | Storage (max) |
| ---------- | ------------------------------ | ------------- |
| Starter    | 8 GB, more with RAM you add    | 40 GB         |
| Growth     | 32 GB                          | 80 GB         |
| Scale      | 32 GB                          | 150 GB        |
| Enterprise | 64 GB                          | 300 GB        |

Each plan also includes a storage total, counted in Mediums at 40 GB each:
Starter 40 GB, Growth 160 GB, Scale 640 GB. Storage you add raises it, at $1 a
month per 20 GB.

Each computer can be resized in place up to these maximums without buying an
add-on: Starter 8 GB RAM / 40 GB disk, Growth 32 GB RAM / 80 GB disk, Scale
32 GB RAM / 150 GB disk. RAM or storage you add raises them. vCPU is shared across an account's
computers, up to 8 vCPU on any one computer; add vCPU to your account to run
more. On every plan, paid ones included, a computer is created or resized on its
own up to XL (6 vCPU, 32 GB). Anything larger, such as 2XL (8 vCPU, 64 GB), is
set up on request: the app shows "Contact us" and the API answers `403` with
code `CONTACT_US_FOR_SIZE`. Enterprise accounts, and accounts whose negotiated
plan goes past XL, create and resize past XL on their own. A computer that
already has a larger size keeps it. A macOS computer runs on its own Mac and is priced by that hardware, so
its cores do not count against the account's vCPU (its RAM and its place in the
computer count still do).

### Is RAM per computer or pooled?

Per computer, with a total for the account. Each computer has its own RAM: it
starts at 8 GB and can grow as far as the plan's RAM total: 8 GB on Starter, 32 GB (XL) on Growth and Scale, more with RAM you add. All your computers
together can use up to 8 GB, 32 GB, or 128 GB, plus any RAM you add. So Scale
runs 16 computers at 8 GB each on its own RAM, or fewer computers with some made
bigger (for example 4 at 32 GB), and 50 with memory and vCPU added (50
Mediums need 400 GB and 100 vCPU), with more by the hour from credit past
that. It is never one 128 GB machine.

## Add-ons (pay-as-you-grow)

Plans are a starting point. Rather than pre-pay for a bigger plan, add the vCPU, memory or storage they need
to the one you have. Each add-on is a quantity item on your existing
subscription: it prorates on purchase and renews with your plan, on one invoice.

### What a computer's worth costs

A computer has no fee of its own. Past the plan's own vCPU and RAM, you add memory and vCPU (and
disk past the plan's per-computer disk). A computer's worth of all three,
bought together (the Windows row and `POST /api/account/capacity` sell it this
way), comes to, per month:

| Size   | RAM   | vCPU | Disk   | Price / mo |
| ------ | ----- | ---- | ------ | ---------- |
| Small  | 4 GB  | 0.5  | 20 GB  | $13.50     |
| Medium | 8 GB  | 2    | 40 GB  | $32        |
| Large  | 16 GB | 4    | 80 GB  | $64        |
| XL     | 32 GB | 6    | 150 GB | $118       |
| 2XL    | 64 GB | 8    | 300 GB | $215       |

The app creates Medium and up, so every computer made in it has at least 2
vCPU, and offers up to XL. 2XL is set up on request on every plan; an
Enterprise account, or one whose negotiated plan goes past XL, creates it on
its own. Small (0.5 vCPU, 4 GB) is offered only through the API; a Small
computer that already exists keeps working and can be resized up.

Each size counts the vCPU it launches with: XL 6 and 2XL 8.

### Components

What you can add to the account. Plan & billing sells **memory, vCPU and
storage**: each goes into your account's pool, which every computer draws from,
so you can run more computers or make them bigger. Added now, prorated; given
back at renewal. A computer's resize links to the same rows (**Room to grow**)
when the account needs more room.

| Add-on   | Price          | Unit               |
| -------- | -------------- | ------------------ |
| RAM      | $2.50 / GB / mo | 1 GB              |
| vCPU     | $2.50 / mo     | 0.5 vCPU           |
| Disk     | $1 / mo        | 20 GB block        |
| Windows  | $15 / mo       | 1 Windows computer |

**Windows is included on Scale.** On any other paid plan you buy it per machine,
as a **Windows computer** under Computers on Settings › Plan & billing: the
computer's own memory, vCPU and disk at the rates above, plus **$15/mo** for the
license. A Windows Medium is $32 + $15 = $47 a month; one with fewer vCPU or
less disk costs less (Windows needs 8 GB of RAM or more). That is the point of
the add-on: a Starter plan with one Linux computer and one Windows Medium costs
$76/mo instead of moving to Scale at $399.

The license is not sold on its own, only with the computer's own room, so each
Windows computer below Scale is paid for even when the plan's own vCPU and RAM
are free: two Windows Mediums on Starter cost $29 + 2 × $47 = $123/mo, and the
plan's own Medium stays free for a Linux computer. From eight Windows Mediums,
Scale ($399) costs less. A Windows computer is billed by the month, on a yearly
plan too.

**Or run Windows by the hour.** On Starter and Growth, a Windows computer can
also run by the hour from credit, with no license to buy: $0.10/hr for a
Medium, $0.20 Large and $0.40 XL while it runs, billed by the second, and only
its disk while it is stopped. New computer shows the rate before you
launch, beside adding Windows to the plan.

The license is a COUNT, not a label on a particular computer: hold three, run
three Windows machines. Creating a Windows computer uses one, and you cannot
give back a license a running Windows machine depends on. The surcharge is flat
per machine rather than a multiplier, because the license is per machine and the
extra RAM and disk Windows needs are roughly constant too. Windows needs 8 GB
RAM or more, so the Small size is Linux-only.

**Mac and iPhone, while they're in stock.** A Mac (a macOS computer on a Mac mini)
starts at **$100/mo** (a Mac mini M4 VM; an M6 VM is $120/mo) and a physical iPhone at **$65/mo** (an iPhone 12). Any account can
get a Mac or an iPhone 12 that is ready now, at that price, from Settings →
Plan & billing → Custom machines or the New computer sheet ("Get one"). A
model with none ready shows "Out of stock", and "Ask us" reaches the team.
The **iPhone 17** (a brand new iPhone 17e or 17) is **$129/mo** and by
request: ask for one there and the team sets it up for you. Once it is
turned on for your account, you get one yourself with "Get one", like the
others. Neither is part of a plan: each is billed on its own, as below.

A Mac or iPhone from Orgo's stock is billed monthly: on the plan's monthly
bill when the plan is monthly (the first, partial period prorated), otherwise
on a monthly bill of its own. Your credits pay its part of each bill first
(see below), then the card on file. You need a card only when your credits
do not cover what is due at checkout.
Each one is kept for at least one month and its payments are not refunded; a
cancel takes effect at the end of a billing period, after that first month.
With no card, a renewal your credits cannot pay ends the hardware, as a
declined card does, and on a plan's bill that can come before the month is up.

Where a model shows a weekly price, you can choose Weekly instead of Monthly
at "Get one". A week is billed on a weekly bill of its own (never on your
plan's bill), renews every week until you cancel, is kept for at least one
week and is not refunded. A cancel takes effect at the end of the paid week.
A month costs less than four weeks.

Self-serve maximums (same on every plan): up to **+800 GB RAM**, **+200 vCPU** and **+4 TB disk** added, and **50 computers** on the plan, with more by the hour from your credit up to **999** an account. Past the added capacity, we raise the cap for your account on Enterprise; nothing else about the account changes.

### Adding capacity through the API

Everything the Plan tab does is also an API, for a platform that gives each of its own users a computer and scales from one to thousands. Authenticate with an account-wide API key (`Authorization: Bearer <key>`); a workspace-scoped key is refused.

- `GET /api/account/capacity`: what the account may run and runs, what it has added, its self-serve cap, how many of each size it may still add, and every size with its monthly price.
- `POST /api/account/capacity` with `{ "size": "medium", "count": 10 }`: adds the memory, vCPU and disk of that many computers to the account's pool on the card on file, at the prices above (up to 500 per request). It adds no computers to the count: every paid plan includes 50, and past 50 a new computer runs by the hour from credit, up to 999 an account. Add `"dry_run": true` to see the charge first. Optional: `"os": "windows"` (adds the license too; by the month only), `"term": "yearly"`.
- Any shape, not only the sizes: `{ "ram_gb": 8, "vcpu": 1, "disk_gb": 40, "count": 100 }`, or a size with parts replaced, `{ "size": "medium", "vcpu": 1, "count": 100 }`. RAM 4, 8, 16, 32 or 64 GB, vCPU 0.5, 1, 2, 4, 6 or 8, disk 20 to 300 GB in steps of 20: the shapes a computer can be created with. Past XL (`"size": "2xl"`, 64 GB or 8 vCPU) is set up on request: `403` with code `CONTACT_US_FOR_SIZE`, dry runs included, unless the account is on Enterprise or a negotiated plan past XL (`past_xl_allowed` in `GET`). Every shape is billed at the component rates below, so 8 GB / 1 vCPU / 40 GB is **$27** a month and 8 GB / 0.5 vCPU / 40 GB is **$24.50**. `GET` returns the rates (`rates`; `computer_usd` is 0) and maximums (`per_computer_max`), so a fleet can price any shape without a dry run.
- Memory, vCPU or storage on their own, into the account's pool: `{ "addon": { "vcpu": 4, "ram_gb": 16, "disk_gb": 100 } }`, amounts to add (vCPU in 0.5 steps, RAM in whole GB, disk in 20 GB steps), at the component rates below: that one is **$65** a month. `"dry_run": true` and `"term": "yearly"` work the same; nothing else goes beside it.

Send an `Idempotency-Key` header (a UUID) with every purchase: a retry with the same key gets the first answer back and never buys twice. Past the self-serve cap the answer is `403` with code `capacity_cap`, naming the limit; ask us and we raise it for the account.

### Monthly or yearly

Add-on capacity is sold on two terms; Plan & billing sells monthly, and
yearly is bought through the API (`"term": "yearly"`). Yearly locks the rate in for a year and
saves **20%**, the same as a yearly plan. It is paid up front and cancels at
the end of the term.

| Size   | Monthly     | Yearly           | Yearly total |
| ------ | ----------- | ---------------- | ------------ |
| Small  | $13.50 / mo | $10.80 / mo      | $129.60 / yr |
| Medium | $32 / mo    | **$25.60 / mo**  | $307.20 / yr |
| Large  | $64 / mo    | $51.20 / mo      | $614.40 / yr |
| XL     | $118 / mo   | $94.40 / mo      | $1,132.80 / yr |
| 2XL    | $215 / mo   | $172 / mo        | $2,064 / yr  |

Yearly rates per component: RAM $24/GB/yr, vCPU $24 per 0.5 vCPU/yr, disk
$9.60 per 20 GB block/yr. A Windows computer is sold by the month only.

Every yearly rate is exactly twelve months less 20%, so the saving is the same
at every size rather than drifting with the shape of the machine.

Add-on capacity is priced above in-plan capacity, and deliberately: a plan is a
commitment and an add-on is not. A Medium inside Starter is $29 a month,
inside Growth $24.75 ($99 for four) and inside Scale $24.94 ($399 for
sixteen). At $32 monthly, or $25.60 committed for a year, an add-on
carries a real premium for staying flexible without being a penalty for it, so
nobody has to size up a whole plan to get one more machine.

There is no daily term. A term is a promise about a machine and capacity here
is pooled, so "which computer is on the daily one?" has no answer. Short-term
use is metered by runtime instead.

**Reducing a mixed holding.** Monthly capacity is given back first; yearly runs
to the end of the term it was paid for.

### How add-on capacity is charged

Add-ons are quantity items on your subscription. Buying mid-cycle is prorated,
so you pay for the days left in the period and the full price from the next
invoice.

What you add goes into your account's totals of memory, vCPU and disk. It is
not tied to a particular machine: any computer you create can use it. It does
not change the computer count, which is 50 on every paid plan.

**Giving capacity back.** Lower the number on its row (Memory, vCPU and
storage) on
Settings › Plan & billing (/settings/billing). The reduction takes effect at your next renewal, not
straight away: the current period is already paid for, so you keep the capacity
and keep being billed for it until the boundary, then both drop together.
Raising the number back before the renewal cancels the reduction. Capacity that
a computer is currently running on cannot be given back; delete the computer
first.

**Usage credits** are your one credit balance. It pays for agent chat, hourly
computers and bills. Every plan
includes a monthly amount, and you can add more at any time.

**Auto top-up** charges your card a set amount whenever your credit falls below
a set amount, at most 5 times in 24 hours (Settings › Plan & billing › Credit).
Adding credit in the app offers it with the purchase, for example "Top up $25
when credit falls below $5"; it is never turned on without that choice.

**Your credits pay every bill first, then your card.** Your whole credit
balance (credit you bought, starting credit and your plan's monthly credit) pays
each Orgo bill before the card, dollar for dollar up to the bill: a new plan,
upgrades (monthly to yearly too), renewals, add-ons, Bops plans, dedicated
computers and hardware. Promotion codes are not combined with credits on a new
plan's first bill: with a code, the code pays toward the first bill and your
credits pay from the first renewal. The card on
file pays the rest, and Checkout always asks for a card for later bills. The
part of a bill paid from credits is not refunded. Credit top-ups and the AI
credit add-on are never paid from credits (they buy credits).

### Per-computer hard cap

No single computer may exceed **8 vCPU · 64 GB RAM · 300 GB disk** - on any plan, with any amount of add-ons. Past XL (6 vCPU, 32 GB) is set up on request, except on Enterprise and negotiated plans that go past it. Retired plans are held to the same per-computer ceiling as the plans on sale today; none of them allow a bigger computer.

## Operating Systems

| Tier       | Linux | Windows                          |
| ---------- | ----- | -------------------------------- |
| Starter    | Yes   | $15/mo a computer, plus its room |
| Growth     | Yes   | $15/mo a computer, plus its room |
| Scale      | Yes   | Included                         |
| Enterprise | Yes   | Included                         |

Below Scale, a Windows computer is its own memory, vCPU and disk plus the $15
license: a Medium is $47 a month, a smaller one less (see
[Add-ons](#add-ons-pay-as-you-grow)). Macs and iPhones
are on every plan while in stock, billed on their own.

## Network

| Tier       | Bandwidth  |
| ---------- | ---------- |
| Starter    | 1 Gbps     |
| Growth     | 1 Gbps     |
| Scale      | 1 Gbps     |
| Enterprise | 1 Gbps     |

## Hosting

Serve an app that runs on a computer at a web address. Hosting is included on
every account, the Free tier too, at no extra charge: nothing is billed per
URL, per deploy or for traffic. Public URLs and custom domains work on Linux
and Windows computers on Orgo's hosts (not a Mac, an iPhone or an Android
phone); apps kept running and Deploy from GitHub are for Linux computers.

| What                                              | Free                    | Starter | Growth | Scale | Enterprise |
| ------------------------------------------------- | ----------------------- | ------- | ------ | ----- | ---------- |
| Public URLs per computer                          | 10                      | 10      | 10     | 10    | 10         |
| Public URLs per account                           | 200                     | 200     | 200    | 200   | 200        |
| New public URL names a day                        | 50                      | 50      | 50     | 50    | 50         |
| Private links (your team only)                    | Yes                     | Yes     | Yes    | Yes   | Yes        |
| Open a link to anyone, or share it by email       | With $5 of bought credit | Yes    | Yes    | Yes   | Yes        |
| Deploy from GitHub                                | With $5 of bought credit | Yes    | Yes    | Yes   | Yes        |
| Deploys per computer                              | 10                      | 10      | 10     | 10    | 10         |
| Apps kept running per computer                    | 20                      | 20      | 20     | 20    | 20         |
| Custom domains                                    | 0                       | 0       | 1      | 3     | No limit   |

- **Public URLs.** A public URL is `https://<name>-<4 characters>.web.orgo.dev`,
  serving a port on the computer. A computer can have up to 10, on every plan;
  the computer's Public URLs tab shows how many are used ("2 of 10") and the
  API says the limit (`limit` in `GET /api/computers/{id}/public-urls`). An
  account can hold 200 across all its computers and make 50 new names a day.
  These limits are the same for every plan and count on the account that
  pays for the workspace.
- **Private by default.** A new public URL opens only for people in the
  workspace, who sign in to Orgo first. The workspace owner or an admin
  chooses who can open each one.
- **Link sharing needs a paid plan or bought credit.** Opening a URL to anyone
  with the link, or sharing it with people by email, needs a paid plan, or at
  least $5 of credit bought with none of it refunded or disputed. Either the
  person or the workspace owner can be the one who pays. A Free account that
  has bought no credit keeps its URLs with its team; starting credit does not
  count. The API answers `403` with code `UPGRADE_REQUIRED`, and the list says
  it first (`can_share`).
- **Deploy from GitHub.** Builds a repository on the computer and serves it at
  a public URL. It needs a paid plan or bought credit, by the same rule
  (`can_deploy` in `GET /api/computers/{id}/deploy-projects`). Up to 10 deploys
  on one computer, on every plan.
- **Custom domains.** Your own domain on a computer, by plan: Starter 0,
  Growth 1, Scale 3, Enterprise no limit, counted across all the workspace
  owner's computers. Free (Pay as you go) has none, and bought credit adds
  none. Plans no longer sold: Hacker and Developer 0, Team and Startup 1,
  Max 3. On a plan with none, the computer's Domains tab names the plans
  that include one.
- An unclaimed agent account can't publish (public URLs, deploys or custom
  domains) until a person claims it.

## Residential IP data

A residential IP sends a Linux computer's Google Chrome out through a home
internet connection in the US, instead of Orgo's data center address. Every paid
plan includes residential data each UTC month:

| Tier          | Included each month |
| ------------- | ------------------- |
| Pay as you go | None                |
| Starter       | 1 GB                |
| Growth        | 5 GB                |
| Scale         | 20 GB               |
| Enterprise    | 20 GB               |

- Included data is used before data you buy. It does not roll over: what is left
  at the end of the month is gone, and the next month starts again at the plan's
  amount. A plan on a trial includes none until the trial ends, and a paused plan
  none while it is paused. If your plan ends, pauses or switches before the 1st,
  the amount follows the new plan from that day.
- More data is **$3 per GB** from your Orgo credits or by card, in whole GB, up to
  20 GB in one purchase. It is added at once and isn't refunded once added.
  Deleting the account, or a ban for abuse, loses what is left.
- An account can add up to 20 GB a day (UTC), every way together: purchases,
  card payments, auto-buy and its plan's data. So one purchase is at most what is
  left of that today. After that, more can be added from 00:00 UTC.
- An account can make 10 purchases a day (UTC). Card payments you start count
  too, even ones you don't finish. After that, buy again from 00:00 UTC.
- A free account can buy it too: pay by card, or add credit and pay from it.
  Starting credit alone can't buy residential data. Once the account has paid by
  card or added credit, its whole balance pays, starting credit included.
- Auto-buy, which adds 1 GB from your credits when data runs low, isn't
  available yet. Until it is, buy data yourself.
- The data is the workspace owner's, and every computer in their workspaces draws
  on it, teammates' too.
- Linux computers only, and US IPs only. To ask about other countries, email
  [contact@orgo.ai](mailto:contact@orgo.ai).
- Ticketing, video and music streaming, gambling, torrent and speed test sites are
  blocked, and so are banks, payment sites and government sites on the network.
- Some sites check the browser as well as the IP. A home IP doesn't get past a
  site's own bot checks or sign in.
- API: `GET /api/billing/residential-data` (what is left, the price, purchases),
  `POST /api/billing/residential-data/buy`, and `POST
  /api/computers/{id}/residential-proxy` to turn it on. With no data left, turning
  it on answers `402` with code `RESIDENTIAL_DATA_NEEDED` and spends nothing.
- Buying spends credits, so an agent asks its human first. Only the workspace
  owner can buy: their own sign in, or a key with Billing write for the whole
  account. A key without it reads `can_buy: false` with `needs_billing` when
  nothing else stops the buy. `402` `INSUFFICIENT_CREDITS` means the credits don't
  cover it: `price_cents` is 1 GB, `total_cents` the whole. `503`
  `RESIDENTIAL_DATA_UNAVAILABLE` means data can't be added right now, by anyone,
  and nothing was charged: paying, adding credit or asking the owner won't help, so
  try later. `429` `rate_limited` with `buy_opens_at` means the day's 10 purchases
  are used: nothing was charged, and `GET /api/billing/residential-data` says
  `buy_blocked_reason` `day_limit` with the same `buy_opens_at`, the next 00:00 UTC.
  Without `buy_opens_at` it is the API's own rate limit: wait `retry_after` seconds.
  `GET /api/billing/residential-data` says the account's limit a day,
  `daily_gb_limit`, and what is left of it today, `daily_gb_left`; `max_gb` is
  never more. `409` `RESIDENTIAL_DATA_LIMITED` with `max_gb` means less than you
  asked for can be added now: buy that much or less. `429`
  `RESIDENTIAL_DAILY_LIMIT` with `buy_opens_at` means the account reached its
  limit of GB for the day: nothing was charged, `buy_blocked_reason` is
  `daily_data_limit`, and more can be added from `buy_opens_at`, the next 00:00
  UTC. At the network's reserve, or the account's limit for the day, this month's
  plan data that hasn't moved yet is `included.waiting_bytes`, not `left_bytes`,
  until it can move. `409`
  `RESIDENTIAL_PURCHASE_IN_PROGRESS` or `202` `RESIDENTIAL_PURCHASE_PENDING` means
  a purchase is still finishing: read `GET /api/billing/residential-data` in a
  minute.

## AI Credits

Monthly credits included with each plan, for AI agent usage, hourly computers
and bills.

| Tier       | Credits    |
| ---------- | ---------- |
| Starter    | $5/mo      |
| Growth     | $10/mo     |
| Scale      | $50/mo     |
| Enterprise | Custom     |

Credits also pay for the model behind an agent on your computers (Hermes). It
works out of the box on Orgo's models, with no API key to set up, at the same
per-token rates as agent chat: the provider's price plus 10%, or your own rate if your account has an AI pricing deal. Put your own
Anthropic key in Secrets and the agent uses that instead, billed by Anthropic.

## Team Members

How many people the owner of a workspace can add to it, besides themselves. Members (share-link guests included) and invites that have not expired both count, per workspace, against the workspace owner's plan.

Teammates are the people who work in your workspaces. The people your agents or your app serve (clients, customers, end users) never need a seat, on any plan.

| Tier       | Members    |
| ---------- | ---------- |
| Free       | 5          |
| Starter    | 5          |
| Growth     | 5          |
| Scale      | 25         |
| Enterprise | No limit   |

- Every plan includes people, Free too: 5 on Free, Starter and Growth, 25 on Scale, no limit on Enterprise. Email invites and share links both count.
- People already in a workspace keep their access when the owner's plan changes. A lower plan only stops adding new people, and invites that are waiting can't be accepted until there is room.
- The workspace Bops keeps its computers in (named "bops") follows the owner's Bops plan instead: 5 people on every Bops plan (an older Bops Max keeps 10). An Orgo plan adds no one there, and a Bops plan adds no one to other workspaces.
- Accounts with dedicated computers or hardware from Orgo can add up to 25 people to their workspaces, whatever their plan.

## Feature Comparison

| Feature                  | Starter    | Growth     | Scale      | Enterprise |
| ------------------------ | ---------- | ---------- | ---------- | ---------- |
| Computers always on      | 1 Medium   | 4 Medium   | 16 Medium  | Custom     |
| Computers (count)        | 50, 999 with credit | 50, 999 with credit | 50, 999 with credit | Up to 999 |
| vCPU included            | 2          | 8          | 36         | Custom     |
| Persistence              | Persistent | Persistent | Persistent | Persistent |
| RAM per computer (base)  | 8 GB       | 8 GB       | 8 GB       | Custom     |
| RAM per computer (max)   | 8 GB       | 32 GB      | 32 GB      | Custom     |
| RAM included (all computers) | 8 GB   | 32 GB      | 128 GB     | Custom     |
| Storage per computer (base) | 40 GB   | 40 GB      | 40 GB      | Custom     |
| Storage per computer (max) | 40 GB    | 80 GB      | 150 GB     | Custom     |
| Operating systems        | Linux; Windows $15/mo + its size | Linux; Windows $15/mo + its size | Linux + Windows | Linux + Windows; Mac and iPhone fleets |
| Network                  | 1 Gbps     | 1 Gbps     | 1 Gbps     | 1 Gbps     |
| Platform credits         | $5/mo      | $10/mo     | $50/mo     | Custom     |
| Residential IP data      | 1 GB/mo    | 5 GB/mo    | 20 GB/mo   | 20 GB/mo   |
| Public URLs per computer | 10         | 10         | 10         | 10         |
| Custom domains           | 0          | 1          | 3          | No limit   |
| People besides you       | 5          | 5          | 25         | No limit   |
| Workspaces               | 9999       | 9999       | 9999       | 9999       |
| Templates                | Provided   | Provided   | Provided + custom | Provided + custom |
| Computer-control API     | Yes        | Yes        | Yes        | Yes        |
| Python SDK + OpenAI API  | Yes        | Yes        | Yes        | Yes        |
| Own OpenAI or Anthropic key | Yes     | Yes        | Yes        | Yes        |
| 1-click cloning          | Yes        | Yes        | Yes        | Yes        |
| Real-time collaborative canvas | -- | Yes | Yes | Yes |
| Support                  | Community  | Priority   | Advanced   | Dedicated (SLA) |

Enterprise "Custom" limits are set per account as part of the contract: the number of computers (optionally per operating system: Linux, Windows, macOS, Android, and physical iPhones), vCPU and RAM per computer, the account RAM pool (checked when a computer is created and when one is resized larger), storage per computer, the most bandwidth a computer can be set to, and whether the account can build templates. Workspaces are not a contract term: every account, Enterprise included, can own up to 9999. A physical iPhone counts as a computer but brings its own hardware, so it does not draw on the account's RAM or vCPU. The limits apply while the subscription is active.

## Capabilities

The following capabilities are included on every paid plan (Starter, Growth, Scale, and Enterprise).

- **Live desktop + computer-control API** - view and drive every computer in real time via browser or programmatic API.
- **Python SDK + OpenAI-compatible API** - drop-in compatible with existing OpenAI clients; no rewrite required.
- **Your own OpenAI or Anthropic key** - on the chat API, `X-OpenAI-Key` pays for `orgo-agent` (Computer Use, on OpenAI's GPT-6.1 Terra) and `X-Anthropic-Key` pays for Claude models and `orgo-agent-opus-5.5`. That provider bills you, and the run uses no AI credits. `orgo-agent` asks OpenAI for its fast tier, so it is billed at OpenAI's fast tier rates. No other provider's key works there. Agents on your computers use the Anthropic key in Secrets (Codex uses the OpenAI key).
- **1-click computer cloning** - duplicate any computer's state into a fresh VM (`Cmd+C` / `Cmd+V` in the dashboard, or `POST /api/computers/[id]/clone`).
- **Discord community + founder office hours** - direct access to the team via the public Discord and scheduled office-hour calls.
- **Isolated cloud computers** - no shared tenancy on paid VMs; each computer is yours alone.
- **Direct SSH (key access)** - real key-based `sshd` access to Linux computers via your own keypair, reachable over the IPv4 high-port fallback (`ssh -p <port> root@<host>`). Linux-only; distinct from the `orgo ssh` WebSocket terminal. Included on every paid plan.

## Collaboration

Inviting people is on every plan, Free too, up to the plan's [team members](#team-members): Free 5, Starter 5, Growth 5, Scale 25, Enterprise no limit.

- **Team workspaces + email invites** - organize computers into shared workspaces and invite teammates by email, with Can chat or Full access.
- **Guest share links** - generate short-lived links that let anyone holding them join as a guest (Can chat or Full access). Each guest counts as a team member.

Included on Growth and Scale (not Starter):

- **Live cursors + multiplayer chat** - see teammates' cursors on shared computers and chat inline alongside the cursor.

## API Access

All plans include API access for programmatic computer management. See the [SDK documentation](https://docs.orgo.ai) for details.

API rate limits are per account, by plan (Enterprise has Scale's unless its contract says otherwise). Extra API keys share their account's limits.

| Limit                                    | Free | Starter | Growth | Scale |
| ---------------------------------------- | ---- | ------- | ------ | ----- |
| New computers a minute                   | 2    | 10      | 30     | 60    |
| New computers an hour                    | 10   | 100     | 300    | 1000  |
| New computers a day                      | 20   | 500     | 2000   | 5000  |
| Starts, stops and other changes a minute | 10   | 60      | 120    | 300   |
| Agent runs a minute                      | 5    | 30      | 60     | 120   |
| Agent runs at once                       | 1    | 5       | 20     | 50    |
| New workspaces an hour                   | 10   | 30      | 60     | 120   |
| Control calls a minute (key or MCP)      | 300  | 1500    | 3000   | 6000  |
| Other key requests a minute              | 120  | 600     | 1200   | 3000  |

Starts, stops and other changes are also limited to 10 a minute per computer, and control calls (screenshots, clicks, typing, commands) to 600 a minute per computer, on every plan. New computers, changes, agent runs and new workspaces count every caller, the dashboard too; control calls and other requests count only API keys, a computer's own key and MCP sign ins. Over a limit, the API answers `429` with code `rate_limited`, the `limit` it hit and `retry_after` in seconds.

## Free Tier

The Free tier is for exploring Orgo: sign up, browse the dashboard, create workspaces, and generate API keys at no cost.

A new account gets **$3 of starting credit** the first time it is used signed in, if it passes the free checks below; if it does not, adding a card gets it. On the Free tier it runs **one Medium Linux computer** (2 vCPU, 8 GB RAM, 40 GB storage) billed by the hour: **$0.055/hr** while it runs and **$0.0027/hr** while it is stopped, so $3 is about 54 running hours. AI usage draws on the same credit. Free computers share a limited number of places; at busy times a launch or Start may ask you to try again in a few minutes. Credit you buy is never held back by that limit.

**The free checks.** A free account with no card is limited per network, per email domain and per day; gets nothing free from a server, datacenter or VPN address or a throwaway inbox; and gets the starting credit once per inbox. When a check holds something back, the API answers `429` with code `NEEDS_CARD` and often a `verify_url` (a browser check or a phone code) that lifts it; adding a card lifts it too. `NEEDS_PAYMENT` means the card already lifted another account: buy credit or a plan. Buying credit or a plan lifts every check. Agents: the full rules, with the error codes, are in the "Try Orgo free" section of /orgo.md.

When the credit runs out, the computer stops. With no plan, its files are kept for **7 days**. To get it back with its files, add credit and start it, or choose a paid plan with room for it (the plan takes the computer over). Adding credit stops the deletion, but the computer stays stopped until you press Start. After 7 days with no plan and no credit, the computer and its files are deleted.

**An older computer on Free.** A computer stopped on a Free account from before it ran by the hour (made on a plan you left) can run by the hour from your credit, so you keep it. Start shows its price and asks first. It keeps its size, and it needs credit for an hour of it and an hour of your other hourly computers; on starting credit alone it is your one computer and takes a free place. Running is the plan's rate plus 25% for its size and its real disk, and never under the [By the hour](#by-the-hour) price of the same vCPU and RAM: 1 vCPU and 4 GB on 20 GB is $0.027/hr, 2 vCPU and 16 GB is $0.087/hr, and a Medium stays $0.055/hr. Stopped is its real disk at the plan's rate. Windows needs an offered size. From there it is an hourly computer like any other: when the credit runs out it stops and its files are kept for 7 days, and a plan with room takes it over. API: `POST /api/computers/{id}/start` with `{"billing_mode":"hourly"}`, by the workspace owner, signed in or with an account-wide key that has Billing write. Without it the `402 upgrade_required` says `hourly_available`, `running_usd_per_hour`, `stopped_usd_per_hour`, `credit_usd` and `credit_floor_usd`, and nothing changes.

Free users **can** also be invited into a paid workspace and create/use computers there, funded by that workspace owner's plan. A Free account can invite up to 5 people into each of its own workspaces. An unclaimed agent account can't invite anyone until a person claims it.

**Hosting on Free.** A Free account gets public URLs like every plan: up to 10 on a computer, each open only to people in its workspace. Opening one to anyone with the link, sharing it by email and Deploy from GitHub need a paid plan or $5 of bought credit, and custom domains start on Growth (see [Hosting](#hosting)).

**Agent accounts.** Where anonymous agent signup is on (`POST /api/v1/agents`, see /auth.md), an AI agent can make its own account with no email. It starts with **$2 of starting credit** (about 36 running hours), on the same Medium Linux computer at the same hourly prices. One computer at a time until it has added **$5 in all**; once it has, up to **5 hourly Medium Linux computers and 5 workspaces**, and the account no longer expires. An account that has added less expires 72 hours after signup unless a person claims it. A claimed account keeps its Free-tier rules; claiming adds no credit.

Accounts created before 2026-05-12 that remain on the Free tier keep any computer they already had (1 vCPU / 4 GB RAM, 8 GB storage, 100 Mbps network); creating new computers requires upgrading.

## Enterprise

For organizations with custom requirements: large fleets (more than 999 computers, or volume pricing past 50), custom compute sizes, dedicated support, an SLA by contract, dedicated hosts by contract, custom privacy and data controls, and implementation sprints run with our team to get your agents into production.

Email [spencer@orgo.ai](mailto:spencer@orgo.ai).

## Definitions

- **Computer**: A virtual desktop (VM) accessible via VNC through browser or API.
- **Persistent**: Computer runs 24/7. State is preserved across restarts.
- **RAM pool**: Total RAM available across all computers in your account: the plan's, plus any RAM you add. Each computer has its own RAM, up to the plan's per-computer maximum; mix and match sizes up to the pool limit. The vCPU pool works the same way.
- **Computer limit**: every computer counts, plan and hourly, running or stopped: 50 on every paid plan. Past 50 a new computer runs by the hour from credit while the credit covers an hour of all hourly computers, the new one included, up to 999 an account. Past 999, contact us.
- **Platform credits**: Monthly allowance for AI agent operations (Anthropic, OpenAI integrations). It also pays for hourly computers and bills.
- **Team member**: A person besides the owner with access to a workspace (Can chat or Full access), including share-link guests and invites that have not expired. Counted per workspace against the owner's plan.
- **Clone**: Duplicate an existing computer's state into a new VM. Counts against the computer limit and the RAM and vCPU pools.
- **Own key (BYOK)**: your own OpenAI or Anthropic API key. It pays that provider for the model instead of your AI credits. The computer is billed as usual.
- **Real-time collaborative canvas**: Live cursors and inline chat on shared computers; teammates see each other's positions and can leave messages without leaving the screen.
- **Guest share link**: A short-lived URL that lets whoever holds it join the workspace as a guest; it can open on a single computer. Each guest counts as a team member.
- **Direct SSH**: Real key-based `sshd` access to a Linux computer using your own SSH keypair (distinct from the `orgo ssh` WebSocket terminal). Connect over the IPv4 high-port fallback.
- **Public URL**: A web address Orgo gives an app running on a computer (`https://<name>-<4 characters>.web.orgo.dev`, serving one port). Up to 10 on a computer on every plan, Free included. New ones open only for the workspace's people; opening one to anyone with the link needs a paid plan or bought credit.
- **Custom domain**: Your own domain serving an app on a computer. Counted across all the workspace owner's computers against the owner's plan: Starter 0, Growth 1, Scale 3, Enterprise no limit.
- **Staging/test usage**: Only production usage is billed.

## Links

- App: [orgo.ai](https://orgo.ai)
- Docs: [docs.orgo.ai](https://docs.orgo.ai)
- Contact: [spencer@orgo.ai](mailto:spencer@orgo.ai)
