Google Cloud brings GA agent observability, audit logging and cost controls to BigQuery and Looker

On November 14, 2026, Google Cloud announced an expansion of its Conversational Analytics agent platform across BigQuery and Looker, adding generally available tracing, audit logging, monitoring dashboards and cost controls, alongside a new open source SDK and evaluation tool, according to a Google Cloud blog post.

Concrete changes shipped

• Conversational Analytics agent telemetry now auto-emits with no setup required (generally available)
• The open source BigQuery Agent Analytics SDK adds one-line instrumentation for ADK or LangGraph agents, streaming tool-invocation trees, token counts and turn latency into BigQuery (generally available)
• The Conversational Analytics API gets always-on admin audit logs plus optional data-read audit logs in Cloud Audit Logs, with support for VPC Service Controls, customer-managed encryption keys (CMEK) and regional or multi-regional data residency (generally available)
• Prebuilt agent monitoring dashboards shipped in BigQuery Studio, and Looker added System Activity dashboards for engagement and token usage (generally available)
• A new big_query_max_billed_bytes setting caps per-query byte scans using dry-run estimates (generally available)
• Prism, a new open source evaluation tool, lets teams build reusable test suites with regex assertions for Conversational Analytics agents grounded in BigQuery tables or Looker Explores (generally available)

What it means for people building with AI

For teams shipping agentic apps on Google Cloud rather than demos, this delivers concrete, free to use tooling to trace, audit, control cost for and evaluate agents built with ADK or third party frameworks like LangGraph. A built in privacy guardrail also fails closed on logging whenever row level security, column policy tags or dynamic data masking apply, so sensitive data stays protected by default.

What's next in preview

Google Cloud also moved two features into preview: daily input and output token quotas configurable per project, user or agent through the Cloud console or the gcloud CLI, and native BigQuery agent evaluations via the Gemini Enterprise Agent Platform eval service and Colab.

Sources

• Enterprise AI agent observability with BigQuery and Looker (Google Cloud, November 14, 2026, official)

Written with AI assistance, strictly from the sources listed above. Before it went live, automated checks verified its dates, sources and every language edition.