Bencher

Bencher provides a continuous benchmarking platform that runs the same benchmark suite on identical bare‑metal hardware both locally and in CI pipelines. It stores each run’s results, lets users query and visualize trends through a web console, and applies configurable analytics to flag performance regressions before code is merged. By using dedicated bare‑metal runners, the system reduces the variance typical of shared CI agents from over 30 % to under 2 %, making observed changes reliable.

The tool is aimed at development teams whose products are sensitive to performance, such as database engines, compilers, browsers, runtimes, networking stacks, cryptographic libraries, and operating systems. It integrates with existing benchmarking frameworks via a CLI and can be invoked from GitHub Actions, enabling automated detection of regressions in pull‑request workflows. Open‑source under the MIT license, Bencher is self‑hostable, offers a free tier, and does not require a subscription, making it suitable for organizations that need reproducible, low‑noise performance testing.