e2e is an end-to-end testing framework for web and mobile apps that uses AI agents to automate test execution based on natural language descriptions of goals.
The framework addresses the problem of writing and maintaining brittle end-to-end tests by letting developers describe what they want to accomplish in natural language, then having an agent drive the application to reach that goal. Tests combine agent-driven steps with locators and assertions to verify results. A key efficiency feature is that once an agent step and its verification are recorded, subsequent test runs replay those actions without making new model calls until the application changes. Tests that contain only assertions and locators require no model at all. The tool supports bringing your own API key, subscription, or local model for the AI component.
The framework targets teams building web and mobile applications who want to reduce test maintenance overhead. It works with Chromium, Firefox, and WebKit through Playwright for web testing, and supports iOS and Android through simulators and emulators. The tool provides optional hosted infrastructure through kernel browsers and EAS simulators, along with a GitHub reporter for posting results as pull request comments. Developers can choose between different model providers and decide whether to use agent-driven steps or traditional assertion-based testing depending on their needs.
The project maintains active documentation that ships with the SDK for offline access, and development discussion happens on Discord. The CLI collects anonymous usage telemetry covering which commands and engines run and where failures occur, but excludes test content, application content, and credentials, with an option to disable collection entirely.
