Description
The worst part of end-to-end testing isn't writing it, it's maintaining it: one button moves and a dozen selectors go red, and there goes another weekend. e2e flips the approach. You state the goal in a sentence, like "upgrade the workspace to the Pro plan", an AI agent clicks, types and works through the flow, and locators and assertions in the same test nail down the result.
It also remembers how it got there. Agent steps that a later assertion verified are recorded, and the next run replays them with zero model calls until the app actually changes. Tests without agent steps need no model at all. Bring a subscription, an API key or a local model, and use the same style for web, iOS and Android.
Natural-language steps: agent.act describes what to do and agent.assert describes what should appear, right next to ordinary expect and getByRole checks in one test.
Record and replay: verified agent actions are saved and replayed on later runs with no model calls, so the agent only steps in again when the UI changes.
Web and mobile: the browser engine runs on Playwright with Chromium, Firefox and WebKit; the mobile engine drives iOS simulators and Android emulators.
Bring your own model: plug in a subscription, an API key or a local model, with decision-model executors for bounded semantic actions and assertions.
One-command setup: npx e2e init asks for engine and provider, then writes a config and an example test; the examples folder ships Vite, Next.js, Expo and SwiftUI projects.
CI and hosted devices: a GitHub reporter posts results as a pull request comment, and Kernel hosted browsers and EAS hosted simulators plug in too.
It also remembers how it got there. Agent steps that a later assertion verified are recorded, and the next run replays them with zero model calls until the app actually changes. Tests without agent steps need no model at all. Bring a subscription, an API key or a local model, and use the same style for web, iOS and Android.
Features
Natural-language steps: agent.act describes what to do and agent.assert describes what should appear, right next to ordinary expect and getByRole checks in one test.
Record and replay: verified agent actions are saved and replayed on later runs with no model calls, so the agent only steps in again when the UI changes.
Web and mobile: the browser engine runs on Playwright with Chromium, Firefox and WebKit; the mobile engine drives iOS simulators and Android emulators.
Bring your own model: plug in a subscription, an API key or a local model, with decision-model executors for bounded semantic actions and assertions.
One-command setup: npx e2e init asks for engine and provider, then writes a config and an example test; the examples folder ships Vite, Next.js, Expo and SwiftUI projects.
CI and hosted devices: a GitHub reporter posts results as a pull request comment, and Kernel hosted browsers and EAS hosted simulators plug in too.
