---
title: "AgentTwin sandbox: test a synthetic workflow"
description: "Try a clean test and a timeout-after-write fault locally without an account or provider credentials."
canonical: "https://agenttwin.si/sandbox/"
---

# AgentTwin sandbox: test a synthetic workflow

The shipped CLI example is a credential-free local test world with synthetic ticket and notification collections. It is a reference workflow, not your application or a hosted production replica. Node.js 22.12+ is required. There is no signup, sales form or cloud-provider charge for this local example.

## Clean baseline

```sh
npx agenttwin example --out ./agenttwin-example
cd agenttwin-example
npx agenttwin run --seed 00000000000000000000000000000001 --fault none
```

The prepared example expects one ticket and one notification. Inspect the reported assertions and recorded final state; an exit code alone does not prove unrelated application behavior.

## Inject a fault

```sh
npx agenttwin run --seed 00000000000000000000000000000001 --fault timeout-after-write
```

This intentionally failing example exposes duplicate writes after a timeout. Use the emitted run ID with the CLI's verify, reset and replay commands. Follow the [full quickstart](/docs/quickstart/) for exact receipt inspection and reset behavior.

## Boundaries

The command executes the generated application locally with host filesystem and network access; review its files first. Synthetic provider tokens are not production credentials. Each run creates fresh synthetic state. This sandbox is not a public unauthenticated cloud execution endpoint. Testing your own application requires reviewed invocation, fixtures, operation mappings and unchanged assertions. Optional cloud execution needs a configured runner and is subject to provider billing and API quotas.