TestDriver MCP

TestDriver automatically runs every pull request in a real desktop sandbox, finds bugs, and builds regression tests.

Hosted MCP Server

npx add-mcp 'https://mcp.testdriver.ai/mcp'

Installs into Claude Code, Codex, Cursor and more

Documentation

Find & fix bugs in flows other tools miss

Go beyond the limitations of traditional automation to find bugs in the end-to-end processes that other tools can't test.

  • No Selectors Required

    Canvas, video, and controls tools can't target.
  • Cross-App Workflows

    Across browsers, windows, popups, and apps.
  • Third-Party Integrations

    OAuth, payments, and software you don't control.
  • AI Apps & MCP Servers

    MCP tools inside Claude, ChatGPT, and more.
  • Desktop Applications

    Windows and Linux apps, not just websites.
  • Plugins & Extensions

    Chrome and VS Code extensions, in the real app.

See what you can automate

Real tests running against real apps. Pick one to see how it was built.

See all Demos

How it works

Tests your app the way a person uses it

TestDriver looks at the screen, finds elements from plain-language descriptions, and controls a real machine with the mouse and keyboard. No selectors.

Step 1 · Provision

Provisions the infrastructure

Every test starts on a fresh cloud sandbox. One provision call launches the app under test on a real machine: Chrome at a URL, a Chrome extension, VS Code, an Electron app, or a downloaded installer. There are no browsers to manage and no test rigs to maintain.

Step 2 · Find

Finds elements with AI vision

Describe an element the way you'd describe it to a person, like 'Sign in button'. TestDriver looks at the screen, locates it, then clicks, types, and scrolls with the real mouse and keyboard. Nothing depends on CSS or XPath selectors, so it reaches what the DOM can't describe.

Step 3 · Assert

Asserts that things work, and finds the bugs

Write a plain-language statement about the screen, such as 'the cart shows one item'. The AI checks it against what is actually rendered and fails the test when it isn't true. That covers charts, layout, spelling, and media that never show up in the DOM.

Upload

Learn More

Assert

“the popup modal shows”

Step 4 · Learn

Learns your app as it runs

After a test passes, TestDriver remembers where it found each element. When the screen looks the same on a later run, it reuses that location instead of asking the AI, so repeat runs are faster, cheaper, and consistent. When the UI changes, it falls back to the AI and learns the new location.

Upload

Learn More

Element cache

Nav

Heading

Image

Button

Footer

Step 5 · Run in CI/CD

Runs in CI/CD to detect regressions

Tests are ordinary Vitest files, so any CI provider can run them with npx vitest run. Run them on every push and pull request, and a regression fails the check before it merges.

Console

Know what broke, why, and whether it’s fixed

Every run lands in one console, with the history, replays, and trends you need to fix failures fast and keep them from coming back.

List of test runs Regression and quality trends dashboard Action log timeline Video replay of a failed run

Platform overview

The whole platform in two minutes

A guided tour of TestDriver showing how tests are written, run on real desktops, and reported on. No selectors, no flaky scripts.

Pricing

Less than the cost of one missed bug.

Starting at $24 per seat per month with 360 testing minutes included.

Pro

Cloud

$24/ month / seat Seats are automatically billed for every GitHub user that gets a PR review or runs a test.

Plus $0.08 Per Testing Minute Testing minutes are billed for the duration the sandbox is running. This happens automatically on PR generation and test runs.

Start Free 14-Day Trial

Review every PR with our AI QA Engineer.

Enterprise

Self-Hosted

Let's Talk

Contact Sales

Unlimited testing with licenses shared across your team.