Google adds records of ai agent work to BigQuery and Looker

Google released new tools to monitor ai agents in its Data Cloud. Data Cloud includes BigQuery, a data warehouse, and Looker, a tool for reports. The tools work for agents that Google made and for agents that teams make with ADK or LangGraph. Those are 2 tools for agent builders. Records of agent work go into tables of the customer. Google gave no test results.

How it works

Agents that Google made send use data to BigQuery with no more work. Teams that make agents with ADK or LangGraph add one line of code. Admins see dashboards in BigQuery Studio and Looker. Each agent owner must start the record of conversations.

The numbers

Each turn of a conversation records 5 token counts and the total time. The system does not record a turn when the data source uses row-level security, column policy tags or masked data. If a record error occurs, the conversation continues.

What is not known

Google gave no test results and no numbers of use. The price for Conversational Analytics is a special price during a trial. No price for after the trial is given. Looker ratings from users are in a test release only.

Insight: Google builds one system for agent work

In November 2026, Google added a catalog of business definitions and a server for cloud rules that agents can read. Then it added records of agent work. Maybe Google wants BigQuery to hold agent facts, rules and records, which makes a change of vendor hard. Leaders must find out where agent records are kept.