Google shares AQuA, an ambient quality agent that diagnoses production AI agents
A Google Developers Blog post dated Oct. 8 introduces AQuA (Ambient Quality Agent), a quality agent that reviews production agent sessions, clusters recurring failure patterns and traces their root causes. The authors note that an agent can return HTTP 200, stay within its latency budget and raise no tool errors while still failing silently in conversation — for example confirming seat 3A without checking availability, or recommending a Florentine steakhouse to a traveler whose profile says vegan. AQuA cross-references failing traces against the deployed source snapshot, uses Gemini 3.8 Flash to anchor a proposed fix to a specific file and line, and exposes the insight payload through agents-cli so a coding agent can apply and replay-verify the fix on a branch.
The post argues that after launch, usage, models, harnesses and tools all shift while infrastructure health checks stay green, so conversation-quality regressions never trigger a deploy warning — the gap this production-side quality tool is meant to cover.
The Outer Loop, Insights First: An Ambient Quality Agent That Diagnoses Your Production Agent - Google Developers Blog
The Outer Loop, Insights First: An Ambient Quality Agent That Diagnoses Your Production Agent
OCT. 8, 2026
Dima Melnyk Product Manager Cloud AI
Elia Secchi Solutions Specialist
#Google#AQuA#Gemini 3.8 Flash#Google Cloud#agents-cli#Gemini CLI#Claude Code#Cursor