OpenComputer is an Apache-2.0 licensed TypeScript platform for building, steering, and deploying serverless agents as code, aimed at developers and small teams who want persistent cloud virtual machines that hibernate and wake in seconds rather than assembling their own agent hosting.

What it is

OpenComputer is a managed platform, a TypeScript CLI, and a project layout that together treat agents as code. An OpenComputer project contains one or more agents that run in the cloud. Running npx @opencomputer/cli init my-agent creates a hello-world project whose opencomputer/ tree holds the cloud agent definitions — opencomputer/project.ts alongside opencomputer/agents/hello-world/agent.ts — plus a package.json. That tree is designed to grow to multiple agents inside a single project. A browser application can live in the same repository, but it keeps its own development and deployment lifecycle separate from the agents.

The concrete problem it solves is the glue between agent code and a running cloud deployment. Deployment is versioned and immutable: npm run deploy -- --alias production publishes an immutable release, and infrastructure details stay behind the OpenComputer API, so the developer edits the agent function, tools, connections, and routing rather than the hosting layer. The specific thing it replaces in the project's own history is the legacy Go oc CLI; the TypeScript opencomputer CLI is focused on managed agents and is separate from that older tool.

Key capabilities

• npx @opencomputer/cli init my-agent scaffolds a hello-world project containing opencomputer/project.ts, opencomputer/agents/hello-world/agent.ts, and package.json.
• npm run deploy -- --watch watches agent code and deploys changes to Development (Cloud) while it runs.
• npm run deploy -- --alias production publishes an immutable, versioned deployment.
• Managed-agent commands include opencomputer whoami, opencomputer agents, opencomputer session "Say hello", and opencomputer run hello-world "Say hello".
• The dashboard project view exposes Agent playground, Sessions, Files, Connections, Channels, Schedules, and Agent schema.
• The CLI targets https://app.opencomputer.dev by default; --api-url or the OPENCOMPUTER_API_URL environment variable point it at another service intentionally.
• Agents run as persistent cloud VMs that hibernate and wake in seconds.

Who uses it and how

• Developers who keep several cloud agents in one repository, expanding the opencomputer/ tree as the project grows.
• Teams that colocate a browser application and agent definitions in the same repository while running the two lifecycles independently.
• First-time deployers who pick an existing project from their account or create a new one on the initial npm run deploy -- --watch, after which that binding is reused automatically.
• Developers testing agents from the dashboard or the CLI while a watched cloud deployment is active.
• Small projects overall: 535 stars, 62 forks, and 30 open issues indicate a young codebase rather than a large established ecosystem.

Getting started

Scaffold with npx @opencomputer/cli init my-agent, run npm install, then npx --package @opencomputer/cli opencomputer login, and deploy with npm run deploy -- --watch. Documentation lives at docs.opencomputer.dev and the dashboard at app.opencomputer.dev.

How it compares

The facts name no other agent platform, so OpenComputer stands alone in this registry among comparable tools. The only adjacent tool named is the legacy Go oc CLI, which this TypeScript CLI is explicitly separate from rather than a replacement marketed against paid products. No licence, hosting, or pricing comparisons to commercial alternatives can be drawn from the available material.

When to use it — and when not to

OpenComputer is a managed service: the README does not document self-hosting, nor any database, object storage, or SMTP component a self-hoster would otherwise have to operate. The README is sparse on pricing, configuration, and the runtime contract of an agent, and the repository carries 30 open issues, so teams needing on-premise control, documented deployment internals, or a stable, widely deployed ecosystem should look elsewhere. Its command surface is TypeScript tooling, which also fixes the language of the workflow.

npx @opencomputer/cli init my-agent
cd my-agent
npm install
npx --package @opencomputer/cli opencomputer login

npm run deploy -- --watch

my-agent/
├── opencomputer/
│   ├── project.ts
│   └── agents/
│       └── hello-world/
│           └── agent.ts
└── package.json

npm run deploy -- --alias production

opencomputer whoami
opencomputer init my-agent
opencomputer agents
opencomputer session "Say hello"
opencomputer deploy --watch
opencomputer deploy --alias production
opencomputer run hello-world "Say hello"