Agent executions
The Agent Executions tab provides a real-time view of the “active thoughts” and processing steps of your AI agents. This is a powerful debugging tool for developers and advanced users.What is an Execution?
An execution is a single instance of an agent’s logic being triggered—usually by a call or a webhook. It tracks the lifecycle of that specific interaction.Key Information Tracked
1. Token Usage
See exactly how many input and output tokens were used for each turn of the conversation.2. Processing Steps
View the timeline of events:- ASR Latency: Length of time to hear and transcribe.
- LLM Thinking: Time taken for the brain to decide on a response.
- TTS Generation: Time to synthesize the audio.
3. Tool Calls & Results
If your agent used a tool (liketransfer_call), you can see the exact parameters sent and the response received from the external system.
4. Raw Logs
For deep debugging, you can access the raw JSON logs of the interaction between Movoice and the underlying providers (Deepgram, OpenAI, etc.).When to check Executions
- Latency Issues: Identify which specific layer (ASR, LLM, or TTS) is causing a delay.
- Logic Failures: If the agent isn’t following instructions, see exactly what prompt was sent to the LLM.
- billing Verification: Audit your token consumption at a granular level.