Pine Computer

Pine Computer is a cloud machine made by Pine AI, the company behind the pine.im assistant. It gives an AI agent a real Chrome browser, a bash shell, a working folder and a desktop that a person can watch. Your product submits a job through an API. The agent opens pages, fills forms, edits spreadsheets and saves receipts, then reports what it produced.

The difference sits in how the machine talks to the model. On a normal PC, an agent takes a screenshot, guesses what changed, and takes another screenshot. Pine Computer sends the model a notice when a page, an app or a file changes. The agent stops re-reading the screen after every step, so a cheaper model can finish more of a long task. Pine's own numbers on SaaS-Bench v1.1 put GPT-5.6 Luna on Pine Computer at a 78.3% checkpoint score for $1.02 in model tokens per task, against 74.3% and $26.50 for Opus 5 with Claude Code. Pine says those runs cost more in its consumer product, and that short tasks show a smaller gap.

This is developer infrastructure, not a consumer app. People reach it through products built on top of it, such as a property management tool that renews rental permits on city portals. The service is in private beta, so access comes through a waitlist. It also runs only in Pine's cloud, which rules out local machines and offline work.

Getting Started

1. Join the waitlist at pinecomputer.io and wait for beta approval.
2. Install a server SDK for Ruby, Go, TypeScript or Python, then create a project key in the console.
3. Create a computer with a country setting and an optional ephemeral flag, then open a session with a browser.
4. Run an agent task in plain language, stream its events, and collect the files it returns.