e2e

Overview

e2e is an open source end-to-end testing framework for web and mobile apps. Tests combine natural-language agent actions with precise locators and deterministic assertions in the same test. Verified agent steps record their actions, and later runs replay them without model calls until the app changes.

Key Features

• Natural-language goals let an AI agent drive the application while tests keep deterministic locators and assertions.
• Verified agent actions are replayed on later runs without model calls until the app changes.
• One shared API supports web and mobile testing.
• Configuration lets teams choose the AI model, agent system prompt, and execution engine.
• Tests without agent steps need no model.
• Web execution supports Chromium, Firefox, and WebKit through Playwright; mobile execution drives iOS simulators and Android emulators through agent-device.

Use Cases

• End-to-end testing for web applications and mobile apps.
• Handling test steps that are hard to express with stable selectors by using an AI agent, while keeping deterministic assertions where precision matters.
• Running tests in CI with cached agent actions replaying when nothing changed.
• Migrating from Playwright, Cypress, Selenium, Detox, or Maestro with migration guides.

Getting Started

• Run npx e2e init; it asks for a web or mobile engine and a model provider, then writes a config and an example test.
• The docs quickstart covers the remaining setup.

Deployment & Requirements

• Tests without agent steps need no model.
• Model access can use the ChatGPT, GitHub Copilot, or SuperGrok subscription, a Vercel AI Gateway or OpenRouter key, or a local server pointed to by the runner.
• Secrets stay in environment variables; the config references named credentials and the runner fills them at run time.
• CI mode switches on automatically with one retry, a read-only trace cache, and replay of verified agent actions without model calls when nothing changed.

Before You Adopt

• License: Apache-2.0. Review its terms before using, modifying, or distributing the project.
• The project is in active development on the way to 1.0.
• Agent-driven tests depend on access to an AI model through a subscription, gateway key, or local server.