Cekura drives text-based conversations against your Vatel AI agent. You export a WebSocket URL and session secret from the agent’s Testing tab, enter them in Cekura, then run an evaluation and review the outcome in Cekura’s results view.
Chat testing exercises the same agent logic you rely on in production—prompts, branching, tool calls, and other “brain” behavior—over a text session. You avoid the cost and setup of full voice (carriers, minutes, STT/TTS) while still validating how the agent thinks and acts.
1. Open Testing on the agent in Vatel
In Vatel AI, open the agent you want to evaluate and go to the Testing tab.
Generate a secret when prompted. You will paste this into Cekura later so Cekura can authenticate to your agent session.
2. Choose the agent, version, and WebSocket URL
Pick the agent and agent version you want Cekura to talk to. If you leave the version empty, Vatel uses the appropriate default for that agent.
Copy the WebSocket URL shown here. Cekura connects to this endpoint for the chat session.
Keep the secret from the previous step handy alongside this URL.
3. Create an agent in Cekura
In Cekura, create a new agent that will represent this evaluation setup.
Open the agent’s configuration and set the channel to chat (or the chat bot mode Cekura exposes for text-based runs).
Under the chat connection settings, choose the custom provider and the WebSocket option. Then paste the WebSocket URL and secret from Vatel into the connection fields.
Save the configuration when you are done.
5. Run a chat evaluation
Start a chat evaluation run from Cekura so it opens a session against your Vatel agent using the URL and secret you configured.
6. View the results
When the run finishes, open the results view in Cekura to see transcripts, outcomes, and any metrics Cekura surfaces for that chat test.
