Now they hand over a whole task, and Agent0 investigates it against your live environment, works out which queries to run, and reports back into the conversation you are already having.
That widens what is worth asking. A direct query answers a known value, such as one service's p95 latency over the last hour. Agent0 takes the open-ended work: why a service got slow, what caused an incident, which dashboards and alerts a renamed metric will break.
Agent0 investigates and reports. It does not change your systems. The only things it produces are dashboards and alerts, proposed for you to review and create yourself, so nothing is written to your code or your tools without you.
Short questions come back in seconds. A real investigation takes longer, so your assistant gets a link, and Agent0 keeps working in the background, then delivers its answer when the run finishes. Open the link to watch the investigation in Dash0 and see which telemetry it used.
Agent0, through the MCP server, consumes Dash0 credits, exactly as a conversation in the Dash0 UI is, whether a message starts a thread or continues one.