Running a Sandbox Scenario

Arga environments let you run agents or application changes in a controlled environment that mirrors real-world behavior.

​

Step 1: Start from your system

Deploy your application or agent into a sandbox environment. This can be done in staging, preview environments, localhost, or a dedicated sandbox.

The sandbox includes your system along with the service twins it depends on.

​

Step 2: Attach the outside world

Connect the twins representing the external services your workflow uses. These twins are compatible with the APIs, MCP interfaces, or CLI tooling used by your application.

Once connected, you can seed the scenario with:

• Mock data
• Data pulled from live production services

This creates a realistic environment where workflows can run.

​

Step 3: Execute the workflow

Run the agent or application logic you want to evaluate. Because the environment includes multiple services, workflows can cross tool boundaries just like they would in production.

During the run, Arga records the full trace of events and actions taken.

​

Step 4: Review the results

After the run completes, you can inspect the trace and observe how the system behaved across services.

This helps identify unexpected behavior, incorrect assumptions about integrations, or failures caused by real-world workflow complexity.