# Komodor Benchmarks

> Open, reproducible benchmarks for AI agents operating cloud-native infrastructure. Each benchmark publishes a leaderboard of agent and model pairs, the metric definitions, the methodology, and the raw result data.

Every benchmark has a page at `/benchmarks/<slug>/`, a Markdown version at `/benchmarks/<slug>.md`, and its full data at `/benchmarks/<slug>.json`. Result files follow the schema at `/schema/result.schema.json`.

## Benchmarks

- No benchmarks are published yet.

## Reference

- [Methodology](https://benchmarks.komodor.com/methodology): how benchmarks are run, reviewed and published
- [Result schema](https://benchmarks.komodor.com/schema/result.schema.json): the shape of one leaderboard entry
- [Benchmark schema](https://benchmarks.komodor.com/schema/benchmark.schema.json): the metadata each benchmark declares