Harness Report is a platform for agent builders who want to evaluate how their agents perform on real tasks. It helps teams connect their code, run evaluations, and review results in one workflow. Users can link a repository, launch an evaluation, and inspect task outcomes alongside model calls and the steps taken along the way. This visibility makes it easier to understand where an agent succeeds and where it may need improvement.

The platform presents evaluation results through task-level reports, with examples showing outcomes such as passing a bug fix, writing tests, and building a feature, while identifying an API update as needing work. Its workflow is organized around connecting a repository, running an evaluation, and exploring the results. Harness Report offers sign-in and a get-started option, and supports continuing with GitHub. Designed for people building and testing AI agents, the service focuses on practical evaluation of agent harnesses, helping users assess performance across varied coding tasks rather than relying on a single summary score.