e2e: Open Source AI Testing Framework for Apps Powered by Natural Language, # 1 on GitHub Trends (6300 + Star)e2e：用自然语言驱动 App 的开源 AI 测试框架，GitHub 趋势日榜第一（6300+ Star）

The writing is intuitive: in test, await app.open ('/settings/billing') opens the page, then await agent.act ('upgrade the workspace to the Pro plan') lets the agent operate, and finally await agent.assert ('the invoice preview shows a prorated amount') verifies the result, and then uses expect to assert the UI state.

The agent is responsible for "how to reach the target", the locator and the assertion are responsible for "the result is correct", and the two sides are written in the same test, which is much more stable than the pure prompt-driven test.

Each step performed by the agent will be recorded, and the next run will be directly played back as is, without calling any models; only when the application interface really changes, the path will be re-planned.

Ordinary tests without agent steps do not require a model at all and can be run locally with zero extra costs on CI.

npx e2e init interactive boot: select the engine (Web or Mobile), select the model service provider, and then automatically generate a configuration file and an example test that can run through.

The official examples directory comes with four independent projects: Vite, Next.js, Expo, and SwiftUI. Each comes with a full set of approved use cases, which can be copied to land on your own technology stack.

Flexible model access: Bring your own subscription, API Key, or local model, and switch freely from cloud to cloud.

The download button provides the complete source code package of the main branch of the project (about 5.1 MB), including the framework source code, four sample projects and all documents; after unzipping, pnpm install & & npx e2e init can be started.

The agent will drive the real browser or simulator to perform operations, please only run in the test environment, do not point to the production environment; accessing the cloud model will incur call costs, and the budget is limited to cut the local model. This article is compiled from self-disclosure materials. The resources are for learning and communication only. Please follow the project Apache-2.0 license.