Agentic Compute Platform

Developer quickstart

Start ACP in minutes. Create a project, get a computer, run a thread, and ship your first working workflow.

Get started
Loading...
Complete platform contract

Use every tenant-facing service through the same versioned HTTP API, TypeScript SDK, and Python SDK—against Computer Agents Cloud or your own local appliance.

43 service groups771 API operationsView the full map

Core concepts

Understand the primitives that define how ACP organizes work, persists state, and executes actions.

View all

Start building

Start and continue threads

Create execution threads, stream progress, and keep work moving inside persistent context

Run project tasks

Create tasks, start task-linked threads, attach comments, and manage review loops from the SDK

Manage environment state

Fork threads and computers, edit messages, inspect diffs, and work directly against persistent environment history

Create agents and teams

Define custom agents, orchestrated teams, models, reasoning effort, and reusable execution behavior

Create webhooks

Trigger ACP workflows from external systems and connect inbound events to threads, projects, and automation

Use system and custom skills

Compose powerful built-in skills with your own custom skills to extend what agents can do in ACP

Work inside projects

Use mission control, plan tickets, assign work to agents, and coordinate execution across one product workspace

Specify computers for execution

Choose the computers agents work on and configure runtimes, GUI access, packages, secrets, and custom Dockerfiles

Publish resources

Ship web apps, cloud functions, auth modules, and agent runtimes from the same ACP project

Test and evaluate agent work

Create deterministic Tests, versioned Evaluations, evidence-bound assurance policies, and auditable release decisions

Queue durable Batch jobs

Put threads, workflows, evaluations, optimization runs, and project tickets on a reorderable capacity-aware queue

Build governed security and evidence systems

Operate Security Agents, scientific evidence review, guardrails, authorization, and release controls from the same API