e2e by TesterArmy MCP Server: Setup for Claude, Cursor and Other Apps
Lets Claude, Cursor and other coding assistants open your web or mobile app, look at it, try locators and run test lines while they write end-to-end tests.
The e2e MCP server lets your coding assistant work inside a live copy of your app. It sees the screen, tries locators, runs test statements and takes screenshots, so each line it writes has already been tried.
It is a command in the e2e testing package from the same team and runs on your computer. Start it from an e2e project, because it reads that project's config. You need the e2e package set up first, since the server has nothing to open without it.
Good for
- Developers who write end-to-end tests with the e2e package and want an assistant to check each line against the live app
- QA engineers who want tests written from what is really on screen
- Mobile teams that test apps in a simulator
Works in
12 should work
Tap an app to see how to set it up and any known problems.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Client does not support this transport.
Client does not support this transport.
Install e2e MCP
Claude Code
- 1
Copy this prompt.
Add the e2e MCP server to Claude Code. Run this command: claude mcp add e2e -- npx e2e mcp Then check that it loaded and tell me if I need to sign in or restart.
- 2
Paste it into Claude Code. It shows the change and asks before running it.
Requires Node.js.
Run it from your e2e project folder, because the server reads the project's e2e config. The docs show npx with the arguments e2e and mcp.
Built from the package data we hold. Not yet tested by us.
Tools
Opens your app in a browser or simulator
Loads your e2e config, starts your app, boots the browser or device simulator and returns the first view of the screen. It can also run one named test first and stop where that test stops.
Try: “Open my web app and show me what is on the login page.”
Runs code ·
open_sessionDoes one thing in the open app
Runs one action from the session's catalog, such as observe, tap, type, navigate, locate, run_code, screenshot or start_recording. Every call checks its arguments first, and a failed action returns the screen as it looks now.
Try: “Tap the Sign in button, then take a screenshot.”
Runs code ·
callLists what the assistant can do in the open app
Returns the catalog of actions for the current session, one line each, or the full details of one action.
Try: “What actions can you take in this session?”
Read ·
toolsCloses the app session
Saves any recording still running, ends the test attempt and stops the app processes this session started.
Try: “Close the session.”
Runs code ·
close_session
From the official docs
Permissions
medium risk · 3 things to watch
Medium risk
The assistant can click, type and navigate in your app and start the app's own process. Point the config at a test environment and keep real customer data out of it.
What it can do
- Read your data
- Create and change things
- Delete things
- Read-only mode
Watch out for
It acts inside your app like a user
The assistant can click, type, upload files and navigate, so it can submit forms or change data in whatever environment the config points to. Point it at a test environment, not production. e2e.tester.army
It starts your app's own process
Opening a session runs the app command from your e2e config. Check that command before you let an assistant start it. e2e.tester.army
Secrets are hidden but screenshots can still show other data
Secret fields are masked and the assistant never sees the plaintext. Everything else on screen, such as emails and names, is visible in observations and screenshots. e2e.tester.army
Security advisories
No published security advisories as of Nov 1, 2026.
Technical details
| Talks to | your app and the sites it loads |
|---|---|
| Sign-in and keys | No sign-in for the server. Test logins come from your e2e project's config and are hidden from the assistant. |
| Data retention | Recordings, screenshots and logs are saved to files in your project. |
Questions
Do I need the e2e package to use it?
Yes. The server is the mcp command of the e2e package and reads your e2e project's config. Without a project there is nothing for it to open.
Which AI apps does it work with?
Any app that can start a local MCP server. The docs name Claude Code and Cursor, and the install box covers more.
Does it work for mobile apps?
Yes. A session can open a web page or a mobile app on a simulator or device. A local device target needs the app's bundle ID in the config, and the server installs no build.
Does the assistant see my passwords?
No. Secret values are replaced by a placeholder in what the assistant reads, and it can fill a secret into a field without seeing the plaintext. Password fields are masked in screenshots.
Can it run several sessions at once?
Yes. The default limit is 4 open sessions and you can raise it up to 16 with a flag. With more than one open, each call must name its session.
Does it need an API key or account?
The server itself needs none. Your own e2e project may need keys for its app or tests, and the docs say variables in the config come from the environment the MCP client started.
Information
updated Nov 2, 2026
- Publisher
- Official
- Package
e2e- Registry
- npm
- Transport
- stdio
- Auth
- None
- Version
- 0.22.0
- License
- Apache-2.0
- Last commit
- Oct 31, 2026
- Open issues
- 35
- Listed in
- Official MCP registry
Sources
Docs, issue trackers and discussions this page is built from. Issues, quotes and findings above link to their own source.
Show all 2Hide
- e2e.tester.army · e2e.tester.army
- e2e.tester.army · e2e.tester.army