Google says Conversational Analytics agents emit telemetry automatically, while agents built with ADK or LangGraph can stream structured telemetry into BigQuery with single-line SDK instrumentation. Queries routed to BigQuery data agents in Gemini Enterprise through agent-to-agent workflows send Cloud Monitoring metrics, Cloud Trace spans and conversation logs to the agent's project.
Operational metrics and sanitized trace spans leave out user prompts, model responses and generated SQL. Agent owners can opt in to per-agent conversation logging, which writes prompts and responses to an agent_events table. Logging is skipped when a knowledge source uses row-level security, column-level policy tags or dynamic data masking.
For cost control, BigQuery can cap bytes scanned per query through big_query_max_billed_bytes, and daily input and output token quotas are in preview. Prism, an open-source tool, runs benchmark suites against Conversational Analytics agents and can write results into BigQuery next to production logs.
Why it matters if you build agents
Teams running agents on BigQuery or Looker can check token use, latency and usage in built-in dashboards without building separate export pipelines. Custom ADK or LangGraph agents can use the BigQueryAgentAnalyticsPlugin to record full tool-invocation trees, which include tool arguments that first-party logging leaves out. Conversation logging is opt-in, and Google says a logging error will not interrupt user chat sessions.
Get stories like this every weekday
One short email with the day's agent tooling news.