This documentation is available as Markdown for AI agents and LLMs. See the full Markdown index or append .md to any documentation URL.

Learn how a coding agent can test your app on a cloud simulator and return evidence for you to review.

Expo Simulators is a limited-access preview. It is not included with paid or free plans and is currently available only to select partners. Join the waitlist if you are interested in trying it.

You may already use a coding agent, such as Claude Code, Codex, or Cursor, on your computer. You ask the agent to change your code. Then you open the app to check that it functions as expected.

An agentic workflow runs a coding agent for you, in the cloud. The workflow starts when something happens in your project, for example, when a teammate opens a pull request (PR). The agent tests the change in your app on a cloud simulator. Then the workflow posts the result on the PR. A person reads the result and decides what to do next.

This section has four guides. Each guide adds one agentic workflow to your project.

How an agentic workflow works

Each agentic workflow has three parts:

• A workflow starts the work. In these guides, EAS Workflows runs the jobs in the cloud. A PR, TestFlight feedback, a GitHub Action, or a manual run starts a workflow.
• A coding agent does the task, such as a test of a change, a new feature, or a bug fix. In a workflow, the agent runs from the command line, without a person.
• An Expo Simulators session gives the agent a running app on a cloud Android Emulator or iOS Simulator. The agent uses the app as a person does and saves evidence.

The following diagram shows one run of an agentic workflow. The highlighted steps run on EAS Workflows and Expo Simulators:

Why the agent needs a simulator

A coding agent can read and change code, but it cannot see if the app functions as expected. To see the app, the agent needs a device that runs it.

With Expo Simulators, your workflows can use Android Emulators and iOS Simulators on EAS cloud infrastructure. A job in the workflow starts a session. Then the agent uses a controller, agent-device or Argent, to tap, type, and scroll in the app.

What the agent returns

The agent does not merge or ship a change. It returns evidence for a person to review:

• A verdict, such as PASS or FAIL, with a report of each check.
• The screenshots that the agent saved.
• The screen recording of the session, on the session page on the EAS dashboard.
• A link that opens the build in a browser preview.

For example, Agentic CI posts this evidence in a comment on the PR:

What the agent can do explains the limits of the agent in these workflows.

Before you start

Agentic workflows build on other Expo tools. If you are new to them, set them up in this order:

1. Create a project and push it to a GitHub repository.
2. Get started with EAS Workflows and run your first workflow.
3. Set up a coding agent on your computer and try it on your project.
4. Check that your Expo account has Expo Simulators access.

Then follow Set up a project for agentic workflows one time for your project. Each guide uses this setup.

Agentic workflow guides

See the following guides for use cases of agentic workflows with Expo Simulators. Each guide adds a workflow and the files of its agent: