Skip to main content
With Simulation mode, you can simulate conversations for your LLM-based agents using predefined scenarios. Use it to test and debug the behavior of your agent in various situations without real user input.

Restrictions

  • Multivariate simulations aren’t supported. This mode runs single simulations only. To compare variants, run a multivariate simulation in the Simulator.

How to use simulation mode

To open simulation mode, click the simulation tab in the interaction panel. The mode is labeled Simulator. In the Select scenario list, select the scenario that you want to simulate. Then click Run simulation. The simulation of the selected scenario starts with the following predefined settings:
  • The simulation runs against the flow that is open in the interaction panel. It uses the Locale and the Snapshot that you selected there. This mode doesn’t support agent targets, so you can’t run a scenario that Cognigy generated from an agent.
  • The simulation runs once. You can’t change this.
  • The simulation uses the Mocking setting of the interaction panel. If mocking is activated, Nodes that have mock code run the mock code instead of their original logic.
  • The simulation uses the default LLM selected on the Large Language Models page if it supports the Simulator. Otherwise, Cognigy uses the platform-provided LLM. If no LLM is available, the simulation doesn’t start.
While the simulation runs, the exchanged messages appear in the interaction panel. To cancel the simulation, click Stop simulation. When the simulation finishes, the interaction panel shows the results. They include whether the persona accomplished the mission, which success criteria it met, and the sentiment.

Troubleshooting

More information

Last modified on November 13, 2026