Testing flows
Use the Playground to run a flow without leaving the canvas.
The Playground runs the flow exactly as it's wired on the canvas, without calling the REST API or saving anything extra — it's the fastest way to check a change.
Opening it
Open the Playground from the canvas toolbar. It shows a chat panel for any flow with a Chat Input/Chat Output pair, and a simpler run/result view for flows built around other input or output types.
Inspecting a run
Each step in the Playground corresponds to one component execution. Click into a step to see:
- The exact value that flowed into that component.
- The exact value it produced.
- Any error raised during that component's execution, attached to the component that raised it rather than surfaced generically.
This is usually faster than reading logs for tracking down which component in a chain produced an unexpected value.
Session history
Playground conversations are scoped to a session. Starting a new session clears conversational memory for components that read it (like an Agent's chat history) without touching the flow itself — useful for testing a fresh conversation without rebuilding anything.