Databricks launches Unity Gateway Insights to find wasted tokens in coding agent traces
Databricks says trace analysis recovered an estimated $1.2M a year in its own MCP costs, and the new tool uses Genie to find similar waste in customers' traces.
What happened
Databricks released Unity Gateway Insights. It reads the records of what AI coding agents did (called traces) and writes a report on where money is wasted. Tokens are the units AI models charge for.
Why it matters to you
Companies that roll out coding agents can pay for failed retries and for expensive models doing simple jobs. A report like this shows where that spending goes.
The takeaway
Databricks now offers a report that points to wasted AI spending in coding agent use.
What it means for you
You can generate a report for the last day, 7 days or 30 days, add up to 10 custom questions, and open the example sessions behind each finding.
The report ranks waste by cost and flags simple tasks that Smart Routing could send to cheaper models.
Databricks is positioning its data platform as the control plane for AI agents, tying gateway, Genie and Unity Catalog together.