Solo.io announced two open source initiatives at KubeCon + CloudNativeCon Europe 2026. The first is agentevals, a project for evaluating and benchmarking agentic AI behavior using OpenTelemetry to capture distributed agent interactions and score them against golden evaluation sets. It supports offline and online modes, trajectory matching, LLM-as-judge scoring, and works with any model emitting OpenTelemetry spans. The second is agentregistry, originally launched in November 2025, which Solo.io is contributing to the CNCF. It provides a centralized registry for cataloguing, discovering, and governing AI agents, MCP tools, and agent skills, with integrations for Kubernetes, AWS AgentCore, and Google Vertex AI.
Table of contents
RelatedSort: