Overview

A run executes the authored conversation, including its scripted turns, simulations, and checks. The GUI starts runs explicitly; saving a test does not schedule or start it.

When to use it

Run a single conversation while investigating a focused change.

Run a single conversation while investigating a focused change. Use the shared batch controls when you need coverage across several saved conversations.

How it works

The editor’s Run action starts the saved conversation and shows its live results.

The editor’s Run action starts the saved conversation and shows its live results. The agent’s responses are generated using the selected environment’s behavior. Inspect the conversation, check outcomes and reasoning, and logs beside the editor.

Run control in the test editor
Run one saved conversation to inspect its outcome beside the editor.

Environment-based testing

For a change to an existing agent, create a development environment from Main, make the change, and add coverage for new behavior. Run the affected tests and existing regression coverage, inspect failures, and repeat after correcting the agent. Once the expectations pass, merge the environment back to Main, or route a portion of traffic to it before merging.

Example

After changing the returns playbook in a development environment, run the return conversation there and inspect its routing and response checks.

After changing the returns playbook in a development environment, run the return conversation there and inspect its routing and response checks. Once that focused run is understood, a batch can exercise other conversations that may also be affected.

Considerations

Account for the selected Environment and credit consumption when running tests.

Runs use the selected environment. Check the environment before interpreting a result or changing the agent. LLM-powered execution consumes credits; see credits and costs.

Next steps

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