Agent Evals

A framework-agnostic Python library for evaluating and benchmarking AI agents. Provides regression tests for known failure modes plus capability scoreboards. Works with OpenAI Agents, LangChain, Strands, or custom agent code.

Source (preview): github.com/boozallen/agent-evals — repo opens once it's released.

Coming soon

Public documentation for this baseline is on the way. It will be published alongside the open-source release.