Skip to content
Pine Computer Request access
Menu

Read: structure.

Ruby 
computer = pine.create_computer(
  location: { country: "US" },
  ephemeral: true
)
session = computer.create_session(browser: true)
session.agent.run("Pull last month's invoices from
  the supplier portals into the ledger.")

  1. Humans see pixels. AI reads what each thing is, and what it can do.

    pixelsstructure

Speed and cost

2–5×

faster1

1/25

the model cost2

Compared with AI agents on ordinary computers, on the same long, real-world tasks across apps.

  1. 1. Pine's preliminary tests. Results vary by task.
  2. 2. Model API cost only.
  3. How we measured

What you can build

Your product hands it a job, and the finished work comes back.

All 26 use cases

Renew permits

Your customer clicks Renew once, and every permit due this quarter is renewed on its own city’s portal.

How it runs
Illustration of an example: in your customer’s property software, three permits are renewed at once, each on a computer of its own working its city’s portal. The portal for 9 Pier Road asks for a sign-in; your customer signs in on its screen and hands it back. Each city’s receipt comes back to its permit’s row as a PDF.

Design your first workflow

Cloud computers for AI agents

What an AI cloud computer provides, when a browser alone is enough, and how to evaluate sessions, saved sign-ins, files and human control.

Human-in-the-loop browser automation

Design human handoff for browser automation: separate viewing from control, pause for sign-in or approval, and verify the result before resuming.

All guides

Build more.
Hand AI the work.

Pine Computer is invite-only while in beta.