Breaking02 / IT
AWS introduces UAEF to unify AI agent evaluation
UAEF converts execution records from different frameworks into a common format so teams can compare results using shared metrics.

Amazon Web Services (AWS) introduced UAEF, its AI agent evaluation tool, on 12 Nov. AWS said it is designed to help development teams using different frameworks compare results before and after changes on the same basis.
Compatibility. AWS said UAEF converts each framework’s execution records into a common format. It evaluates those records using 36 built-in metrics across 7 areas, including tool calling and response quality.
AWS released UAEF as open source under the Apache 2.0 licence. Its introduction confirms that the Python library and web interface use the same evaluation engine.
According to AWS, agents under evaluation can run locally. Enabling result storage uses Amazon DynamoDB and Amazon S3. Evaluations that use a language model as a judge call Amazon Bedrock.
AWS said existing evaluation tools tend to depend on individual frameworks’ execution record formats, creating work to rebuild them when frameworks change. UAEF is designed to compare stored results against a baseline run and identify metrics where quality has declined.
AWS’s introduction does not disclose a release number or details of compatibility changes affecting existing code.