Skip to main content
Arga separates the system you are testing from the external services it acts on. You keep running your app or agent in your own local, staging, or preview environment. Arga supplies realistic service twins and a browser runner that can exercise a reachable URL.

​
The workflow

1

Create the service state

Start a Twin Run with the services you need. Seed it from a saved Scenario, a natural-language description, or an empty baseline.
2

Connect your software

Replace provider base URLs and credentials with the values returned by the Twin Run. Calls now change twin state instead of production state.
3

Run the test

Give a Test Run a reachable application URL and a task. The browser runner executes the flow and records events, screenshots, and assertions.
4

Save what worked

Save a completed flow as a test. You can edit its blocks and parameters, rerun it against a URL, or associate it with a repository for PR Test Runs.
5

Reset and repeat

Reset the twins to their seeded baseline and run the same test again. This keeps provider state controlled across attempts.

​
Core building blocks

Twin Runs

Short-lived or persistent service environments with provider-compatible APIs and interfaces.

Scenarios

Saved seed configurations for one or more twins.

Test Runs

Browser execution against a URL with live events, evidence, and editable blocks.

PR Test Runs

GitHub-triggered execution of relevant generated tests or selected saved tests.

​
What repeatability means

Arga controls the external-service starting state and the browser steps. It does not capture every function call in your application or reconstruct the application’s process state. To reproduce a case, use the same Scenario, reset the twins, and rerun the same Saved Test against the same application build.

​
Evidence and iteration

A Test Run keeps the browser events, generated blocks, screenshots, and final summary together. If a flow needs adjustment, edit its blocks or parameters and rerun it. Saving the result turns the flow into a reusable test definition instead of a one-off prompt.