aicomputer

TypeScript SDK for Agent Computer.

Install

pnpm add aicomputer

Authenticate

import { AgentComputer } from "aicomputer";

const client = new AgentComputer({
  apiKey: process.env.AGENTCOMPUTER_API_KEY!,
});

Quickstart

import { AgentComputer } from "aicomputer";

const client = new AgentComputer({
  apiKey: process.env.AGENTCOMPUTER_API_KEY!,
});

const computer = await client.computers.create({
  displayName: "Quickstart computer",
});

await computer.ensureRunning();

const result = await computer.commands.run(["node", "--version"]);
console.log(result.stdout.trim());

await computer.ensureStopped();

Run a command

const result = await computer.commands.run(["python3", "-c", "print('hello')"]);
console.log(result.exitCode, result.stdout);

For long-lived commands, use the typed command handle:

const server = await computer.commands.start(
  ["python3", "-m", "http.server", "3000"],
  {
    onStdout: (chunk) => process.stdout.write(chunk),
    onStderr: (chunk) => process.stderr.write(chunk),
  },
);

await server.wait({ rejectOnNonZero: false });

Publish a port

await computer.waitForPort(3000);
const published = await computer.ports.publish({ port: 3000, name: "app" });
console.log(published.public_url);

PTY sessions

const shell = await computer.pty.create({
  cols: 120,
  rows: 32,
  onData: (chunk) => process.stdout.write(chunk),
});

await shell.sendInput("uname -a\n");
await shell.resize(140, 40);
await shell.kill();

Docs

• Web docs: https://agentcomputer.ai/docs/typescript-sdk
• Generated API reference is published from the package source into the docs site.