discovered 03 Aug 2026
humanbound
→ View on GitHubHumanbound is an open-source adversarial testing engine designed specifically for AI agents, enabling users to simulate realistic user interactions and potential attacks through live endpoints and multi-turn conversations. Its notable features include the ability to transform test failures into deployable firewall rules and compatibility with both local environments and the Humanbound Platform, making it straightforward to initiate tests without authentication. The tool also supports various integration options, allowing for flexible deployment and configuration.