Data and Agent Lineage

Agent Observability now connects the data behind every agent. Monte Carlo resolves and maps the data objects an agent depends on directly from its tool calls — in this iteration, by parsing SQL references out of agent traces.

A new Lineage tab on the agent page shows the full graph of upstream data dependencies. Active alerts on upstream data are surfaced on the lineage nodes themselves, so your team can spot health concerns at a glance.

The connection works in both directions: a data object's lineage also surfaces any agents downstream of it. When a table goes stale, you can see which agents are at risk. When an agent misbehaves, there is a direct path back to the source data behind it.