Agent Observability in BigQuery and Looker

Logs and dashboards that show what AI agents did, what data they touched, and what they cost, for data and platform teams on Google Cloud.

Conversational Analytics agents send telemetry automatically, and custom agents built with ADK or LangGraph can stream it to BigQuery with a single line of SDK setup
Prompts, responses, and per-turn details are logged to a table you own, with access controlled by your dataset permissions, and logging is skipped when a data source uses row-level security, column policy tags, or data masking
Each logged turn records input, output, thought, cached, and total tokens plus latency, and per-query byte caps block oversized queries
Managed conversation logging, BigQuery Studio agent monitoring dashboards, and Looker System Activity dashboards are generally available; user-rating review in Looker, daily token quotas, and native agent evaluations are in preview
Prism, an open-source tool, runs test suites against Conversational Analytics agents and can stream results into BigQuery next to production logs

Tracking an AI agent's reasoning, data access, and cost across BigQuery and Looker

You run conversational analytics agents or custom agents and need to see what they are doing and spending

Data and platform teams responsible for AI agents on Google Cloud

Join the newsletter to get ahead with AI.

I share strategies, new modules, free guides, and tools to get you better at using AI in your work.