Cloud computers for AI automation agencies: run, bill, and hand off client agents on Orgo
How AI automation agencies give each client an isolated Orgo workspace and scoped key, bill them monthly under their own brand, ship identical agent setups with templates, and transfer the finished build.
TL;DR
Orgo for AI automation agencies means giving each client their own isolated Orgo workspace and scoped API key from your own app, running their agents on full cloud computers, and then either covering the cost, billing the client monthly under your brand, or transferring the whole workspace to them when the build is done (Invoicing guide). The client never signs up to Orgo while you run it for them. The payment email goes out under your business name, replies come to you, checkout carries your business name, logo, and color, and your name is on the client's card statement. Invoicing is on every paid plan. The plan Orgo sells to agencies is Growth: $99 a month or $948 a year as of November 2026, with 4 Medium computers always on (or mixed sizes up to XL) and 5 people per workspace besides you (pricing). Clients you serve never take a seat.
What does an AI automation agency need from a computer layer?
An AI automation agency builds agents that do real work for a client: filling forms in a legacy portal, reconciling invoices in a desktop app, or running a browser plus native tools all day. Those agents need a real computer, not only a browser tab. Orgo gives an agent a full computer with a desktop, a browser, and a terminal, reachable over HTTP and VNC (llms.txt). Orgo is not the agent. You bring Claude Computer Use, OpenAI's CUA, Hermes Agent, OpenClaw, or your own loop (llms.txt).
For an agency, the hard parts are not only "can the agent click." They are:
- keeping one client's computers and credentials away from every other client,
- getting paid for the computers and AI each client uses,
- handing the finished build to the client without rebuilding it on their account, and
- shipping the same setup to client number ten as fast as client number one.
Orgo covers each of those with documented primitives: the Clients API, client billing, client transfer, and templates (Invoicing guide, Templates).
Definition: what is an Orgo client?
An Orgo client is a workspace you create for one of your customers with POST /v1/accounts, plus an API key that can reach only that workspace (Add a client). The key is shown once in the create response, so you store it with your user record. With it, your app can create and control computers, run the Computer Use API, and read files for that client. It cannot reach your other workspaces, and it cannot create workspaces, keys, or clients (Invoicing guide). The client's workspace also stays out of your own workspace list.
You can set optional guardrails on each client: external_id (your id for them, made from name if you leave it out), max_computers, and monthly_ai_cap_usd (Add a client).
How do agencies usually set up client computers, and how does Orgo compare?
| Approach | Client isolation | Who pays | Branding on the bill | Handoff to client | Repeatable setup |
|---|---|---|---|---|---|
| Orgo client per customer | Own workspace and scoped key; key cannot reach your other workspaces (Invoicing) | You, or the client monthly via a payment link (Invoicing) | Your business name, logo, color at checkout; your name on the card statement (Invoicing) | POST /v1/accounts/{id}/transfer; build keeps working (Transfer) | Templates build a golden snapshot once (Templates) |
| One shared agency account for all clients | Separation is up to your own naming and discipline | You, then you invoice separately | Whatever invoicing tool you add | Manual rebuild or credential juggling | Whatever scripts you maintain |
| Client signs up and gives you access | Client owns it from day one | Client pays the vendor directly | Vendor's brand, not yours | Already theirs | Depends on you rebuilding inside their account |
| Generic VPS per client | You set up the desktop, VNC, and network isolation yourself | You, then you invoice separately | Whatever invoicing tool you add | Move or reprovision the server | Base image plus setup scripts that drift (Templates) |
Fair-compare note: the non-Orgo rows describe common patterns, not a specific vendor. Orgo's own docs position it as not a general cloud platform; it runs one thing, persistent computers for AI agents (llms.txt).
How do I add a client and keep them isolated?
- Use your account-wide API key. A client's own key, or any key scoped to one workspace, gets
403on client management (Add a client). - Create the client:
curl -X POST https://www.orgo.ai/api/v1/accounts \
-H "Authorization: Bearer $ORGO_API_KEY" \
-H "Content-Type: application/json" \
-d '{"name": "Acme Dental", "external_id": "acme-dental"}'- Store the returned
api_keywith that client in your own system. It is shown once. - Optionally cap them with
max_computersandmonthly_ai_cap_usd.PATCH /v1/accounts/{id}changes them later, andnullclears a limit. - If a key leaks,
POST /v1/accounts/{id}/keyissues a new one and every earlier key stops working (Replace a client's key).
Reusing an external_id returns 409 ACCOUNT_EXISTS, even after that client is closed (Add a client).
How do I bill a client for their AI workers under my own brand?
Set up payouts once in Settings → Invoicing (bank account, business name, logo, color), then send a payment link (Invoicing guide):
curl -X POST https://www.orgo.ai/api/v1/accounts/$CLIENT_ID/billing \
-H "Authorization: Bearer $ORGO_API_KEY" \
-H "Content-Type: application/json" \
-d '{"email": "billing@acme.com", "price_usd": 250, "ai_usd": 40, "setup_fee_usd": 2500}'price_usdis what they pay each month. It must at least cover their computers, or the request fails withSet a price of at least $N, your cost.(Bill a client).ai_usdis optional AI they pay for at cost, as its own line. Each paid month adds it to your credits and caps their AI at it.setup_fee_usdis charged once, with the first month. Orgo takes no share of it, so it is yours less Stripe's card fee.- The email goes out under your business name, and replies come to you (Invoicing guide).
- While they pay, their computers do not count against your plan. The workspace is held to their price instead: a new computer that would push its cost past what they pay gets
403CLIENT_PRICE_EXCEEDED, so raise the price first. If a payment fails for good, they go back on your plan and you get an email.
Later, through the same Clients API (Invoicing guide):
| Action | Endpoint | What happens |
|---|---|---|
| Change the price or AI amount | PATCH /v1/accounts/{id}/billing | Applies from their next invoice |
| Resend a waiting link | POST /v1/accounts/{id}/billing/resend | Emails the unpaid link again |
| Refund a payment | POST /v1/accounts/{id}/billing/payments/{payment_id}/refund | Full refund from your side; Orgo returns its fee and takes back the AI credits that month added (Refund a payment) |
| Stop billing | DELETE /v1/accounts/{id}/billing | A paid month runs to its end (?now=true ends it now); an unpaid link stops at once (Stop billing) |
How much does the agency keep?
Orgo keeps the client's computer cost and any AI amount. Stripe's card fee (2.9% + 30¢) comes out of your side. The rest is yours (Invoicing guide). The computer cost is each computer's monthly rate on its own, for example $32 for a Medium (2 vCPU, 8 GB RAM, 40 GB disk) (Computer prices).
Orgo's own worked example: a client who pays $250 a month for computers that cost $50 sends Orgo $50, Stripe $7.55, and you $192.45 (Invoicing guide).
| Line (docs example) | Amount |
|---|---|
| Client pays you monthly | $250.00 |
| Orgo keeps computer cost | $50.00 |
| Stripe card fee (2.9% + 30¢) | $7.55 |
| Agency keeps | $192.45 |
This is the docs example, not customer data. Your margin is your price minus those two lines, so it depends only on what you charge and which computers the client runs.
How do I hand the finished build to the client?
When the build is done, send a transfer (Transfer a client):
curl -X POST https://www.orgo.ai/api/v1/accounts/$CLIENT_ID/transfer \
-H "Authorization: Bearer $ORGO_API_KEY" \
-H "Content-Type: application/json" \
-d '{"email": "it@acme.com"}'Once the client accepts while signed in to their own Orgo account (or a new one), the workspace, its computers, and its key are theirs, on their plan and AI credits. Whatever you built on the key keeps working, and they can replace the key later. You stay on as an admin until they remove you, and any billing through you ends, because they now pay Orgo directly (Transfer a client).
- The link works for 14 days. After that, accepting it answers
410TRANSFER_EXPIRED. Sending again replaces the old link. - If their plan can't run the computers yet, they pick one before accepting.
- A computer's VNC password can change when they accept. Fetch the password per session rather than storing it.
DELETE /v1/accounts/{id}/transferwithdraws a link that hasn't been accepted.
That gives an agency two clean business models on one setup: keep running the agent as a managed service and bill monthly, or deliver it as a project and transfer ownership.
How do I ship the same agent setup to every client?
Use a template. A template is a declarative spec for an Orgo computer: hardware, software to install, services to run, required secrets, and what the user sees on connect. Orgo builds it once into a golden snapshot, and every launch starts from it in seconds, identical every time (Templates). Secrets let a template require API keys without baking them in, which matters when each client brings their own credentials (Secrets). One catch: a launch that injects secrets cold-boots from the snapshot's disk instead of restoring its memory, so time that path before you promise a client a launch speed.
| Base image + setup script | Template | |
|---|---|---|
| Time to ready | Minutes of install on every boot | Seconds, restoring a prebuilt snapshot |
| Reproducibility | Drifts as packages and scripts change | Content-addressed, identical every launch |
| Sharing | Copy the script around | Publish a ref, launch it into any of your workspaces |
Source: Templates.
Plan note: any paid plan, Growth included, can launch Orgo's curated templates (Coding, OpenClaw, Hermes Agent). Publishing and building your own templates needs Scale or Enterprise on your own account; below that the API answers 403 UPGRADE_REQUIRED (Templates).
Can my client watch their AI worker inside my app?
Yes. Every computer streams over one WebSocket URL, and any websockify-compatible VNC client, such as noVNC, can embed it in your own app (Embed computers). Keep your sk_ API key on your server and send only the per-computer password to the browser. That password is root on that computer, so fetch it per session from your backend and gate the route with your own auth. A client-side viewOnly flag does not reduce what the password can do (Embed computers).
If you want the server to enforce watch-only screens, the client needs an Orgo sign-in: invite them into the workspace with Can chat access, and Orgo's proxy filters their input. Despite the name, they can't chat with the workspace's agents either. That takes Full access, which also lets them operate the computers (Workspaces, members and clients). Each person you invite takes a seat in that workspace; clients you serve only through the Clients API never do.
What happens when an agency grows past a handful of clients?
An account on a paid plan can have 50 computers, plan and hourly together, running or stopped. That is a count, not what a plan runs: Growth runs 4 Mediums always on. Past 50, each new computer runs by the hour from your credit, up to 999 an account. Past 999, ask Orgo (Scale to thousands).
Plan computers share the plan's vCPU and RAM pool. Past that pool, a create that names no billing_mode, from a key with full access or Billing write (a client's key included), runs the computer by the hour from your credit on its own. Send "billing_mode": "plan" to get a refusal instead (Create computer). To keep computers on the plan, add memory, vCPU and disk from code with the capacity API. Buying capacity adds room, not computers. Use an account-wide key, send an Idempotency-Key on every purchase so a retry never buys twice, and price it first with "dry_run": true:
curl -X POST https://www.orgo.ai/api/account/capacity \
-H "Authorization: Bearer $ORGO_API_KEY" \
-H "Idempotency-Key: $(uuidgen)" \
-H "Content-Type: application/json" \
-d '{"size": "medium", "count": 2, "dry_run": true}'One request buys 1 to 500 computers' worth. On your own, an account can add up to 800 GB of RAM, 200 vCPU and 4 TB of disk. Past that the API answers 403 capacity_cap, and Orgo raises it for your account on request, at the same prices (Account limits).
Which Orgo plan fits an agency?
Growth is the plan Orgo pitches to agencies ("For agencies. An agent for each client."). Invoicing itself is on every paid plan. As of November 2026 (pricing, Account limits):
| Growth | Scale | |
|---|---|---|
| Price | $99/mo or $948/yr | $399/mo or $3,828/yr |
| Always on | 4 Mediums, or mixed sizes up to XL | 16 Mediums, or mixed sizes up to XL |
| vCPU / RAM pool | 8 vCPU / 32 GB | 36 vCPU / 128 GB |
| Disk per computer, before add-ons | 80 GB | 150 GB |
| Computers per account | 50, up to 999 by the hour from credit | 50, up to 999 by the hour from credit |
| People per workspace, besides you | 5 | 25 |
| AI & Platform credits | $10 a month | $50 a month |
| Windows | $47/mo per computer ($15 license plus a Medium's capacity) | Included |
| Templates | Launch curated ones | Also publish and build your own |
| Support | Priority | Advanced |
Two details matter for agencies. First, a Windows computer on Growth is $47 a month in all, a $15 license plus a Medium's capacity; without one the create gets 403 WINDOWS_REQUIRES_SCALE. A workspace your client pays for is the exception, because the license is part of the computer cost their price already covers. Second, Orgo's own pricing math says that from 14 Medium computers, Scale costs no more than Growth plus add-ons (pricing.md).
What errors will I hit?
| Error | When | Fix |
|---|---|---|
403 "Use your account-wide API key to manage accounts" | A client's key or any workspace-scoped key on /v1/accounts | Use your account-wide key |
403 UPGRADE_REQUIRED | Adding a client without a paid plan | Move to any paid plan |
409 "Connect a bank account first, so you can be paid." | Billing before payouts are set up | Finish Settings → Invoicing |
400 "Set a price of at least $N, your cost." | price_usd below the client's computer cost | Raise the price |
409 "Your client is already paying for this workspace." | A second payment link for a paying client | PATCH the price instead |
403 CLIENT_PRICE_EXCEEDED | A new computer would cost more than the client pays | Raise their price first |
410 TRANSFER_EXPIRED | Transfer link older than 14 days | Send the transfer again |
403 capacity_cap | Capacity purchase past the self-serve maximums | Ask Orgo to raise it |
403 WINDOWS_REQUIRES_SCALE | Windows on Starter or Growth without a license | Buy a Windows computer, or move to Scale |
FAQ
Does my client need an Orgo account? Not while you run it for them. They get a workspace and key through you and never sign up to Orgo (Invoicing guide). They only sign in to Orgo if you transfer the workspace to them or invite them into it (Transfer a client).
What does my client see when they pay? Your brand. Checkout carries your business name, logo, and color, and your name is on their card statement (Invoicing guide). The payment email goes out under your business name, and replies come to you.
Do my clients count as seats? No. The people your agents or app serve never need a seat on any plan. Seats are for people you invite into your workspaces: 5 per workspace besides you on Growth (pricing).
Can one client's agent reach another client's computers? A client key reaches only that client's workspace. It cannot reach your other workspaces or create workspaces, keys, or clients (Invoicing guide).
Can I cap a client's AI spend?
Yes, for AI billed through Orgo. Set monthly_ai_cap_usd when you add the client, or bill ai_usd, which caps their AI at that amount each paid month (Add a client, Bill a client). A model key of their own isn't capped.
Does the build break when I transfer it? No. Whatever you built on the key keeps working after the client accepts (Transfer a client). Anything that stored a computer's VNC password should fetch it again, because it can change at the handover.
Which agent framework should an agency use? Any. Orgo provides the computer, not the agent. Claude Computer Use, OpenAI's CUA, Hermes Agent, OpenClaw, or your own loop all work (llms.txt).
Related reading
- Give every customer its own AI agent computer
- Templates for vertical AI agents
- Scoped API keys for AI agents
- Share an AI agent computer with your team
- What is computer use?
- Cloud desktops vs headless browsers
Sources
- Orgo docs index (llms.txt)
- Invoicing guide
- Add a client
- Bill a client
- Refund a payment
- Stop billing
- Transfer a client
- Replace a client's key
- Templates
- Secrets
- Embed computers
- Workspaces, members and clients
- Scale to thousands
- Create computer
- Account limits
- Computer prices
- Orgo pricing and pricing.md