Build with Pine Computer.
Explore Pine Computer's capabilities, then request beta access to build your integration.
Start
- Pine Computer overview A cloud computer for AI agents, with a browser, a shell, files and a desktop a person can take over.
- How a Computer works One persistent Computer, many sessions working on it, and credentials that stay on your server.
- Getting startedBeta Create your first Computer with a server SDK, then drive its browser, run a shell command and collect its files.
Guides
- Sessions The unit of work on a Computer, with its own browser window, shell and files, sharing the Computer's sign-ins.
- Agent Run agent work by turn, answer its questions from any process, and build an action-and-screen timeline with the server SDKs.
- Skills Turn a workflow the agent completed, or one a person demonstrated, into a reviewed skill that later runs follow.
- Files and shell A session's working directory, the files you give the agent, the files it returns, its browser downloads, and its shell.
- Artifacts The files a Computer keeps beyond its sessions, which the agent saves and you read and write with the project key, even while the Computer is stopped.
- Web SDKBeta Embed a Computer's desktop with @pinesandbox/computer-web, let a person take control, and handle reconnects.
- Persistence What a Computer saves, what it restores, and which files your application must keep.
- Tokens and scopesBeta Which credentials stay on your server, and the one token a browser receives.
- IntegrationBeta Install the SDKs, set up credentials, and run Pine Computer from your backend and your web app.
Production
- Best practicesBeta How to run Pine Computer reliably in production, from credentials and the Computer lifecycle to agent work, the desktop, usage and support.
- Troubleshooting What a sign-in, install, Computer, agent or desktop problem means, and how to fix it.
- Limits The quotas each project has, the limits each Computer has, and what happens when you reach one.