create-simulation

Agent simulations are created by building a simulation suite tied to your agent model, defining test cases with prompts and success criteria, then running the suite to collect structured results. This process validates agent behavior without real deployments.

Running simulation test suites requires an active Synthflow API key and internet access. You use these to orchestrate the simulation environment, execute the defined test cases, and retrieve performance results for evaluation.

Yes, suite management allows you to create, update, and organize multiple simulation cases within a single suite. Each suite is tied to a specific agent model and contains cases with defined prompts and success criteria.

You evaluate AI agent performance by defining cases with specific prompts and run configurations, then executing a simulation. The simulation gathers structured results based on your success criteria to analyze agent behavior and coverage.

The best way to define success criteria is during case definition, where you specify prompts and run configurations for each test case. These criteria are then used to evaluate the structured results gathered after running the simulation.

A key limitation is the strict requirement for an active Synthflow API key and continuous internet access to run simulations. Additionally, simulations evaluate behavior without real deployments, meaning production environment variables may differ.