Watch it think: debug in the Playground
Goal. Use the live Playground to see an agent's step-by-step reasoning and tool choices, so you can find and fix where a flow goes wrong.
Ask your agent a question that should trigger a tool (e.g. a calculation or a document lookup), then read the Playground's step-by-step trace to confirm it chose the right tool.
The Playground is interactive — it shows the agent reasoning and choosing tools in real time, not just the final answer.
- 1Open the Playground and send a message. Watch the agent's reasoning and the tool it picks unfold step by step — this is your window into why it answered the way it did.
- 2When an answer is wrong, the trace tells you where: did it skip retrieval, pick the wrong tool, or get a bad input? Fix that specific block instead of guessing.
- 3Iterate: tweak a block or its Role/instructions, re-run in the Playground, and confirm the reasoning now goes the way you intended before you ship anything.
You'll see. A live, step-by-step trace of the agent thinking and selecting tools — turning a black box into something you can debug.
Cost. Free; Playground runs use only the model usage of each test message you send.
Takeaway. The Playground makes the agent's reasoning visible, so debugging is targeted — you fix the step that failed, not the whole flow.