Add TestDriver to GitHub

Get an AI code review on every pull request, backed by a real run of your app.

Install the TestDriver GitHub app, and the TestDriver agent reviews every pull request in your repository.

• It reads the diff for bugs.
• When a user would see the change, it runs your app in a real sandbox to prove what it finds.
• It writes and maintains end-to-end tests whenever you ask.

There’s nothing to install locally.

Before you begin

• A GitHub repository where you can install apps (you own it, or you’re an admin of its organization).
• A TestDriver account. You create one or sign in during installation.
• A paid plan or trial for automatic reviews. On the free plan, the agent still reviews a pull request when you @mention it. See pricing.

1. Install the GitHub app

   Open TestDriver on GitHub. Install the app on the repositories you want reviewed, and sign in to TestDriver when prompted.

   You can add more repositories later from your GitHub settings.

2. Get your first tests

   Choose a repository to set up. TestDriver opens a Set up TestDriver issue in it.

   The agent picks up the issue, writes a few sample tests for your app, and runs them. It opens a pull request with the tests and their recordings. Review it and merge it like any other pull request.

3. Open a pull request

   From now on, every pull request you open gets a review without anyone asking. The agent reads the change and runs it when there’s something to drive.

   It posts a single review comment with:

   • A summary
   • A walkthrough of the changed files
   • Any bugs it found, each labelled observed (seen during a run) or suspected (inferred from the code)

   Pushing new commits triggers a fresh review.

   The review is a comment, so it never approves or blocks a merge. To fail a pull request when a test breaks, run your tests in CI.

4. Ask for more with @testdriverai

   Mention the agent on any issue or pull request to ask a follow-up question or request a new test:

   @testdriverai Write a test that logs in, adds the first product to the cart, and confirms the cart badge shows 1.
   

   The agent replies in the thread and posts screenshots as it works. When it’s done, it opens a pull request with the test.

   Describe what a user does and what they should see, not how the page is built. “Confirm the cart badge shows 1” works better than “check #cart-count”.

Next steps

How the agent works

What the agent does on pull requests, issues, and mentions.

Prevent regressions

Run tests in CI and have the agent fix them when your app changes.

Run tests in CI/CD

Run your TestDriver tests in GitHub Actions.

Set up a project

Write and run tests yourself in your own repository.