Overview

Tests specify user messages and checks on agent turns. Simulations describe a user scenario and an outcome, then let AI generate that part of the conversation. Both exercise your agent in the selected environment.

Choose the conversation style

Use scripted turns when a known input or sequence must produce a particular response, route, or tool call.

Use scripted turns when a known input or sequence must produce a particular response, route, or tool call. Use a simulation when the user can take different paths toward the same outcome. A test conversation can contain both.

Compare behavior

Compare messages, checks, conversation length and starting context for Tests and Simulations.

The same support scenario

For an order-status agent, a scripted test can send “Where is order 123?” and check that the order lookup tool was called.

For an order-status agent, a scripted test can send “Where is order 123?” and check that the order lookup tool was called. A simulation can describe a customer who has lost their order number and passes only when the agent provides an appropriate next step. The second example tests the outcome across generated messages rather than fixing every user reply.

Shared editor and saved conversations

Open Tests and New test to author either approach in the shared editor.

Open Tests and New test to author either approach in the shared editor. A saved conversation can combine scripted and simulated segments. Its selected launch event and starting persona or variable values apply to both styles.

Explore each approach

Compare the Test editor with the Simulation editor.

Use the up and down arrow keys to select a result, Enter to open it, and Escape to close the search.