Hercules

Testing a web application usually means maintaining code that breaks every time the interface moves. Hercules takes plain Gherkin steps instead and runs them as end to end tests, so the person who writes the scenario no longer has to be the person who can automate it.

It drives the browser through Playwright and installs from a PyPI package or a Docker image. You point it at a feature file, a test data directory and an output path, or hand it a project base and it expects those folders inside. Every run leaves a JUnit XML file and an HTML report, which is what a pipeline needs to fail a build on.

The agent covers more than clicking through a page.

• API testing: endpoints can be exercised alongside the interface inside the same feature file.
• Security checks: the same Gherkin flow carries security assertions end to end.
• Visual validation: vision capabilities compare snapshots and components against what the scenario expects.
• Multilingual content: test cases written in other languages drive pages in those languages.
• Python sandbox: a step can call your own script with full Playwright access, for selector strategies or data processing that a sentence cannot express.

Hercules is written in Python and released under AGPL-3.0, and the community contributes extra tools to the agent. It suits teams that already describe behavior in Gherkin and want those files to run, rather than sit beside a separate automation suite that someone has to keep alive.