RepoTrials lets teams benchmark AI coding agents against their own Git history
RepoTrials is a new open-source, local-first tool released under the Apache 2.0 license that converts existing Git commit history into repeatable benchmarks for evaluating AI coding agents. Rather than relying on public leaderboards, engineering teams can test which agent, model, or prompt configuration performs best on their own codebase. The tool reconstructs historical moments when a test was failing and packages them as sealed tasks, then grades agents by checking whether the relevant tests pass without breaking existing ones. Task data, hidden tests, and run artifacts remain stored locally, keeping proprietary code private while the engine stays open. Version 0.1.0 is now available on GitHub and currently works best with Python repositories that have reproducible historical tests.
This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)
Log in to join the discussion and vote.
Log in