Google Cloud's Agent Clinic: build an automated eval suite for a LangGraph agent in 60 min

LangChain · x · 2026-10-01

In Episode 3 of Google Cloud Tech's Agent Clinic, Dani Zamora and mattferoz demonstrate building an automated eval suite for a docs agent built with LangChain/LangGraph, with Merge API Gateway as the intelligence provider.

Key point: terminal test runs won't catch multi-turn agent regressions. The video lays out a 4-step framework for going from vibes to a benchmark for any AI agent. The author also ran the agent on three OSS coding tools: t3dotcodes, opencode, and pidotdev.

Original post →

More from coding & agent

coding & agent channel →