E2e

What e2e does

e2e MCP server gives coding agents a way to work with applications during end-to-end test authoring. The workflow combines natural-language agent actions with conventional locators and assertions, so an agent can drive an app and then verify a specific result in the same test.

The framework targets both web and mobile apps. Web runs can use Chromium, Firefox, or WebKit through Playwright. Mobile runs use iOS simulators and Android emulators through agent-device. The project also lists hosted and device-cloud integrations for several web and mobile environments.

A typical test opens an application route, asks the agent to perform a task such as upgrading a plan, checks a semantic result, and then uses a locator assertion to verify the visible state. The MCP workflow is useful when the coding agent needs to explore the application rather than rely only on already-known selectors.

How it works

Agent steps are expressed as natural-language actions and assertions. The agent can inspect the application, experiment with locators, execute individual test lines, and capture screenshots while developing the test. An agent action that is later verified by an assertion records its actions; subsequent runs replay those actions without model calls until the application changes.

Tests that contain no agent steps do not require a model. For agent-driven steps, e2e supports a user-provided subscription, API key, or local model. The project describes the package as including the SDK, test runner, and CLI, with separate packages for web and mobile engines.

The repository includes standalone examples for Vite, Next.js, Astro, Angular, Expo, SwiftUI, Jetpack Compose, Kotlin Multiplatform, Flutter, and .NET MAUI. These examples demonstrate completed setups with passing suites.

Setup and configuration

Start the setup flow with:

npx e2e init

The initializer asks whether the project targets web or mobile and which model provider to use. It then creates configuration and an example test. The framework documentation is also shipped inside the e2e package, allowing coding agents to read the pages from node_modules/e2e/docs.

Telemetry is enabled by the CLI and sends anonymous usage information such as commands, engines, and failure locations. It excludes test content, application content, and credentials. Disable it with:

npx e2e telemetry disable

or set E2E_TELEMETRY_DISABLED=1.

Tools and capabilities

• Inspect web and mobile applications while writing tests.
• Try and validate locators against the application UI.
• Execute individual test lines through an agent.
• Capture screenshots during test development.
• Combine natural-language actions with locators and assertions.
• Replay recorded agent actions without model calls when the app has not changed.

Limitations and notes

e2e is in active development, and APIs and configuration may change between minor releases. Agent-driven steps require access to a subscription, API key, or local model, while tests without agent steps do not require a model. The available runtime depends on the selected web or mobile engine and its configured environment.