e2e is an agentic end-to-end testing framework for web and mobile apps that lets you describe goals in natural language and verify them with normal test assertions.
e2e is an end-to-end testing framework for web and mobile apps. It combines natural-language agent steps with locators and assertions in the same test, so a test can both drive the app toward a goal and check the result. The README also shows that tests can be written without any agent steps when no model is needed.
The project addresses the friction of writing and maintaining UI tests for web and mobile apps when some flows are easier to express as goals than as low-level scripted steps. It also aims to reduce repeated model usage by replaying recorded agent actions on later runs until the app changes, while still allowing conventional assertions to verify outcomes.
At a high level, a test opens the app, asks an agent to perform a natural-language task, and then checks the resulting state with locators and assertions. If an agent step is later verified by an assertion, the framework records that action and can replay it on the next run without model calls until the app changes. The README says the framework supports a web engine through Playwright-based browsers and a mobile engine through simulators and emulators, and that users can bring their own subscription, API key, or local model.
It is gaining attention because it sits at the intersection of AI agents and practical testing for both web and mobile, a combination that is currently compelling for teams trying to simplify test authoring. The README also highlights ready-made examples across multiple stacks, offline-readable docs, and a quick-start flow, which makes it easy to explore. Its current momentum is reflected in the repository’s large star count and especially the sharp increase in stars this week, though the README itself does not explain that growth.
The README explicitly positions e2e alongside Playwright for browser automation, and it uses Playwright-based engines for web. More broadly, comparable approaches include conventional end-to-end testing frameworks for web and mobile, but the README does not name direct competitors beyond its Playwright-based web engine and its mobile simulators/emulators. It also mentions decision-model executors as part of its package ecosystem for bounded semantic actions and assertions.
AI-explained · grounded in each repo's README