h5i
An agent-native workspace for red-teaming web applications that combines a fast browser with direct control over HTTP traffic, so agents can navigate, capture, inspect, modify, replay and compare requests in one sandboxed, auditable interface.
h5i is a lightweight workspace, written in Rust, built for AI agents doing red-team and web-security testing. It combines browser automation with direct control over HTTP traffic, so an agent can navigate an application while capturing, inspecting, modifying, replaying and comparing requests through one interface rather than a separate proxy and browser.
A session combines page state, a cookie jar, network policy and a request record. Sandboxing, network policies and credential isolation keep autonomous testing contained and reproducible, and every session is logged for auditing. The project states its capabilities should be used only against systems the user owns or is authorized to test.
Its agent-facing interface ships as a skill bundled with the binary, and optional web-security, reconnaissance and testing plugins can be installed alongside it. Confirmed attack flows can be committed to a repository and replayed in continuous integration, with example GitHub Actions workflows for security-regression testing.