e2e
OfficialGives a coding agent a live session on an Android app: the agent sees the screen, acts on it, gets back the test line that repeats each action, and runs the test suite.
Gives a coding agent a live session on an iOS app: the agent sees the screen, acts on it, gets back the test line that repeats each action, and runs the test suite.
e2e
e2e is an end-to-end testing framework for web and mobile apps. Describe a goal in natural language and an agent drives the app to reach it. Check the result with locators and assertions in the same test.
// tests/checkout.e2e.ts
import { test, expect } from 'e2e';
test('a member upgrades to Pro', async ({ app, agent, screen }) => {
await app.open('/settings/billing');
await agent.act('upgrade the workspace to the Pro plan');
await agent.assert('the invoice preview shows a prorated amount');
await expect(screen.getByRole('status')).toContainText('Pro');
});An agent step that a later assertion verifies records its actions, and the next run replays them with no model calls until the app changes. Tests without agent steps need no model. Bring your own subscription, API key, or local model.
Quick start
npx e2e initinit asks for an engine, web or mobile, and a model provider, then writes a
config and an example test. The
quickstart covers the rest.
To see a finished setup in your stack, open
examples/:
Vite, Next.js, Astro, Angular, Expo, SwiftUI, Jetpack Compose, Kotlin Multiplatform,
Flutter, and .NET MAUI, each a standalone project with a passing suite.
Related MCP server: agentlane
MCP server
npx e2e mcp is a Model Context Protocol
server over stdio. It gives a coding agent such as Claude Code, Codex, or
Cursor a live session on your app, on web, iOS, or Android: the agent sees
the screen, acts on it, gets back the test line that repeats each action, and
runs your suite. The server needs no model.
{
"mcpServers": {
"e2e": { "command": "npx", "args": ["e2e", "mcp"] }
}
}See the MCP server docs for setup and every tool.
Packages
Core
Package | What it does |
The SDK, runner, and CLI. | |
Browser engine: Chromium, Firefox, and WebKit through Playwright. | |
iOS and Android engine: simulators and emulators through agent-device. | |
Decision-model executors for bounded semantic actions and assertions, and | |
Runs agent steps on your own coding agent (Claude Code or Codex) over the Agent Client Protocol. |
Integrations
Package | What it does |
Cloudflare Browser Run hosted Chromium for the web engine. | |
Kernel hosted browsers for the web engine. | |
Lightpanda, the open-source headless browser, started per run for the web engine. | |
Local Moli browsers with on-demand layout and screenshots, started per worker slot for the web engine. | |
smol machines browsers for the web engine: by default, each attempt branches a warm Chromium microVM. | |
TestMu AI (formerly LambdaTest) hosted Android emulators, iOS simulators, and real devices for the mobile engine, and hosted Chrome and Edge for the web engine. | |
E2B sandboxes: Android 16 for the mobile engine, and Chromium for the web engine. | |
BrowserStack App Automate real Android and iOS devices for the mobile engine. | |
Limrun hosted iOS simulators and Android emulators for mobile tests. | |
Expo Simulators hosted iOS simulators and Android emulators for the mobile engine. | |
Reporter that posts results as a pull request comment. | |
Reporter that writes the run as one HTML page, with steps, screenshots, and video. |
Community
See Community packages for third-party integrations and how to add yours.
Documentation
e2e.com/docs. The e2e package ships
every page, so coding agents can read them offline in node_modules/e2e/docs.
Join us
Join us in changing how software gets tested. We are hiring.
Questions? Ask on Discord. Issues are for bugs and feature requests.
Want to hack on e2e? See Contributing. We prefer an issue or a Discord discussion before a pull request.
Security
Please don't open public issues for security vulnerabilities. Follow SECURITY.md and report them to security@e2e.com.
Telemetry
The CLI sends anonymous usage data, such as which commands and engines run and
where runs fail, but no test content, app content, or credentials. Opt out with
npx e2e telemetry disable or E2E_TELEMETRY_DISABLED=1.
Telemetry lists every field.
Status
e2e is in active development on the way to 1.0. APIs and config can still change between minor releases.
Available Tools
5 toolscallA
Run one tool of a session by name, with its arguments as an object: call {tool: "tap", args: {target: "n42"}, session: ""}. The session's catalog (from open_session or tools) names the tools and their arguments. Actions report what changed on screen and the test line that repeats them; node ids are valid only for the newest observation.
Inputs: tool (required), args, session.
close_sessionA
Close a live session: save a recording still running, end the attempt, dispose the engine, and stop the app processes the session started.
Inputs: session.
open_sessionA
Open a live session on one target of an e2e project: loads the config, starts the app command the target declares (if any), boots the engine (a browser, a simulator), opens the app (its URL, or the pinned app on a device), and returns the session id, the catalog of tools it can run, and the first observation. Where the project has no e2e config file, pass engine to open a session on that engine: {engine: "@e2e-dev/web", app: {url: "http://localhost:3000"}}. With test, it runs that one test first and stops on the screen where the test passed or failed, with the failure and its line ahead of the catalog. Several sessions can be open at once, one per agent, each with its own engine: pass the returned session id to every later call. Then act with call and look with call {tool: "observe"}.
Inputs: app, test, config, engine, +2 more.
run_testsA
Run tests as npx e2e run does and return what it printed: each test, the failures with their code frames, and the summary. Without files it runs every test the config discovers; files narrows to a test file, a directory, a glob, or one test (tests/checkout.e2e.ts:12). The run is the CLI's own, in its own process, with its workers, retries, replay cache, and report (.e2e/report.json by default); it needs no open session and leaves open sessions alone. One run at a time; cancelling the call interrupts it as Ctrl-C would. To fix a failing test, open_session {test: ":"} stops on the screen where it failed.
Inputs: tag, grep, files, config, +13 more.
toolsARead-only
List the tools a session can run through call: observe, the grammar its engine honors (one tool per action the engine declares, type_secret when a secret is configured, and screenshot and the point tools), locate, run_code (test statements run on the live app), test_lines (the test line each action printed and the statements run_code passed, as a draft test body), start_recording and stop_recording when the engine records video, and the project's own tools. With tool, shows that tool's full description and the JSON Schema of its arguments.
Inputs: tool, session.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 5 tools
Each tool has a distinct primary role: open/close a session, run tests, list session tools, or dispatch one session tool call. There is some overlap between run_tests and open_session with a test argument, and between tools and call, but the descriptions make the boundaries clear.
Three tools follow a verb_noun pattern (open_session, close_session, run_tests), but tools is a bare noun and call is a bare verb. The set is readable and uses lowercase snake_case throughout, but the convention is not consistently applied.
Five tools is well-scoped for a meta-server that delegates most session actions through call. Each tool has a clear purpose and none feels redundant or excessive.
The surface covers the core e2e lifecycle: opening and closing sessions, running test suites, inspecting session tools, and executing them. Minor gaps include no explicit list_sessions or session status tool, though observation is available through call.
Maintenance
Related MCP Connectors
Drive real devices from your AI Coding tool. Embed a client SDK (Unity, Godot, Flutter, iOS/macOS, Android, React Native, Web) in your app, then capture screenshots, traverse the UI tree, inject taps and key events, and run automated test tasks on the physical device over a secure relay.
Order human-run real-device app tests from your coding agent; verified bugs return as fix prompts.
Run multi-step tasks in a real Chrome browser: persistent environments, live view, human takeover.
Agents that test your web and mobile app like real users, on every pull request.
Related MCP Servers
- AlicenseNot gradedqualityAmaintenanceEnables an agent and a human to drive the same live mobile and desktop browser sessions of web/responsive UIs, returning contrast ratios, undersized touch targets, overflow and theme-drift problems as text findings attached to exact selectors instead of screenshot guesses. The agent can also read notes the human pins to a frame, so both act on the same measured screen.95 npmMIT

agentlaneofficial
AlicenseNot gradedqualityBmaintenanceEnables AI coding agents to drive and inspect isolated app instances across web, iOS, and Android, with tools for reading screens, running interactions, and capturing screenshots or evidence.510 npm2MIT- AlicenseNot gradedqualityAmaintenanceEnables AI agents to drive live, URL-synced mobile and desktop browser sessions and receive UI problems as measured text — contrast ratios, undersized touch targets, overflow, occlusion, and theme drift — each attached to the exact selector to fix. Humans and agents share the same live session, so pinned notes and their screenshots can be read back by the agent.95 npmMIT
- AlicenseNot gradedqualityAmaintenanceEnables AI agents to open live mobile and desktop web sessions, measure the rendered UI — contrast ratios, touch-target sizes, overflow, theme drift — and receive findings as text tied to exact selectors, while a human watches and pins notes in the same shared session.95 npmMIT