Solo.io launches agentevals to solve agentic AI’s “biggest unsolved problem”

Solo.io announced agentevals, an open-source framework for evaluating agentic AI, at KubeCon Europe.

Agentevals integrates with Gloo Platform and Envoy Proxy to simulate multi-step infrastructure tasks and produce reproducible logs and metrics.

The project uses OpenTelemetry, supports commercial APIs and open LLMs like Llama 3, and is available on GitHub under Apache 2.0.

Solo.io also donated agentregistry to the CNCF and plans collaborations to expand tests and integrations.