Use Cases

What is an ORB computer?

A computer on ORB is a full isolated Linux environment — filesystem, network, bash, packages. Think of it like a VPS, but managed via API and designed for AI workloads.

Your company creates computers for your users. Each user gets their own computer. Inside that computer, you deploy whatever you want — a coding agent, a research tool, a sales workflow, a custom Python script. ORB doesn't care what runs inside. It just provides the computer.

How it works

You are the ORB customer. Your users are your customers. You use the ORB API to manage computers on their behalf.

Your company (ORB customer)
  │
  ├── User A → Computer A (their agent runs here)
  ├── User B → Computer B (their agent runs here)
  ├── User C → Computer C (their agent runs here)
  └── ... 1,000 users → 1,000 computers

You track the mapping between your users and their computers in your own database. Your users never interact with ORB directly — they use your product.

When User A signs up for your platform, your backend calls the ORB API to create a computer. When User A's agent needs to do work, it runs inside that computer. When User A cancels, you delete the computer.

Is ORB right for you?

ORB is a good fit if:

• You're building a product where each user gets an AI agent. The agent needs a persistent environment — not just an API call, but files, packages, running processes. Each user's agent should be isolated from every other user's agent.

• Your agents are stateful. They maintain context across sessions. They write files, install packages, clone repos, run servers. Restarting from scratch every time isn't an option.

• You need to scale to hundreds or thousands of users. Each user needs their own computer. Running a VM per user gets expensive fast.

• You want API-driven infrastructure. Your backend manages computers programmatically. Create, configure, deploy, execute commands, destroy — all via REST API.

ORB is NOT the right fit if:

• You need stateless function execution (use Lambda, Modal, or similar).
• You're running one agent for yourself (just use a VPS).
• Your agents don't need a filesystem or persistent state.

Example: Marketing agent SaaS

You're building a marketing platform. Each customer gets an AI marketing agent that researches their industry, writes content, and publishes across channels.

1. Customer signs up on your platform.

2. Your backend calls ORB API: create a computer for this customer.

3. Your backend deploys your marketing agent into that computer.

4. The agent runs — it has its own filesystem, its own network, its own packages. It calls LLM APIs, writes drafts to disk, publishes content.

5. Customer logs into your dashboard — you fetch results from the agent's computer via the exec or files endpoint.

6. Customer churns — you delete the computer.

The computer is the customer's workspace. Your agent runs inside it. You control everything via the API. The customer never sees ORB.

Example: Coding agent platform

You're building a platform where developers get an AI coding agent that works on their repos.

1. Developer connects their GitHub repo.

2. Your backend creates a computer, clones the repo inside, installs dependencies.

3. Your backend deploys a coding agent (Claude Code, Aider, your own) into the computer.

4. The agent works — edits files, runs tests, creates branches. All inside its own computer.

5. Developer accesses results through your UI. You expose the computer's port for a live preview, or use the exec endpoint to fetch status.

6. Task complete — computer stays alive for the next task, or you destroy it.

One computer per developer. Each developer's code is isolated from every other developer's code.

Who uses this pattern

Coding agent platforms

Cognition (Devin), Factory AI, Cosine (Genie), Replit Agent — each user gets an autonomous coding agent that needs its own dev environment with repo, build tools, and running servers.

Agent orchestrators

Composio, CrewAI, LangChain — platforms where multi-agent crews collaborate on tasks. Each crew needs a persistent shared environment.

Sales and GTM agents

11x.ai, Artisan, Relevance AI — each customer gets an always-on sales agent that manages prospect pipelines across days and weeks.

Legal and research agents

Harvey AI — agents that process thousands of documents for due diligence and contract analysis over hours-long sessions.

Customer support agents

Decagon — agents that handle live conversations, maintaining context and customer history across sessions.

Economics