Skip to content
Elooo CRMHelp Center Go to website

Agent studio

How to test an agent in Agent Studio before publishing it

In AI Agents › Agent Studio, open the agent and click the Test tab (top right) › Start. Chat with the agent in the emulator; the timeline next to it shows every step, tool and response time. Click a step to jump to the node responsible, fix it, Reset and test again.

Updated on

Before you start

  • Have the agent created in AI Agents › Agent Studio and saved at least once (the go-to-node button only appears on saved agents).
  • Have two or three typical customer conversations handy to use in the test.
Where to find it

AI Agents › Agent Studio › (open the agent) › Test › Start

Step by step

Open the test

  1. Open AI Agents › Agent Studio and open the agent.
  2. Click the Test tab in the top-right corner.
  3. If an orange Missing Required Fields warning appears, check the node and field it lists, fill them in and save. The warning doesn't block testing, but it prevents confusing errors.
  4. Click Start. The chat emulator opens with the execution timeline beside it.

Chat and read the timeline

  1. Type in the emulator the way a customer would.
  2. With each reply, the timeline shows cards with the node name, the model used, input and output tokens and execution time.
  3. Red latency badges flag slow steps; green ones are fast.
  4. Hover over a card to see that step's raw input and output.
  5. For longer tasks, the task list shows which task is active and which are done.

Fix and retry

  1. Click a card's purple transition label: the matching node is highlighted on the agent canvas.
  2. Fix that node's instructions, variables or tools and save.
  3. Click Reset in the emulator to clear the conversation and the timeline.
  4. Run the same conversation again and confirm the reply changed the way you wanted.
  5. Once you're happy, publish the agent as usual.

Tips

  • Test the hard and off-topic questions, not just the easy ones: that's where agents usually fail.
  • The download icon at the top of the timeline saves a file with the full execution, handy when asking for help.
  • For voice agents, the same log appears in the Test tab after a test call ends.

Common problems

  • I don't see the go-to-node button → The agent hasn't been saved yet. Save it and go back to the test.
  • The reply takes too long → Look for the card with the red badge: it's usually a long prompt, a call to an external system or a tool. Simplify that step.
  • The agent ignores a rule I wrote → Check in the timeline which node the conversation went through; the rule may sit in a node the conversation never reaches.
  • The test conversation "remembers" the previous one → Click Reset before each new scenario.

Frequently asked questions

Does testing cost more than using the agent for real?

No. The timeline only shows data the model already returns; it adds no cost.

Can I test through the API instead of typing in the emulator?

Yes. You can trigger the agent via API and the timeline records the steps the same way.

Was this helpful?