Selenium Autopilot

Selenium Autopilot helps you create and run browser tests for your web application. It creates a test plan from your written requirements, generates pytest tests, and runs them in Chrome. It does not host or change your application.

What you need

• Windows, Visual Studio Code, and its Python extension
• Python 3.10 or newer and Google Chrome
• Git and uv
• An OpenAI-compatible AI provider for creating plans and tests (a local provider is also supported)

Download and open in VS Code

Download the ZIP from the GitHub repository using Code > Download ZIP, then extract it. In VS Code, choose File > Open Folder and select the extracted project folder.

Or clone it in a terminal:

git clone https://github.com/adibkhaled/selenium-mcp-server.git
code selenium-mcp-server

In the VS Code terminal, install the project dependencies:

uv sync

Create and run tests for your web app

1. Write down the page URL and the user actions and results you want to test. Include acceptance criteria and use a test environment, not production. Do not include real passwords or other secrets. You can start with the sample requirements.

2. Create a local .env file from the example:

   Copy-Item .env.example .env
   

   Set AUTOPILOT_LLM_API_KEY in .env to your provider key and keep the file private; .env is excluded from Git. The default OpenAI-compatible provider uses https://api.openai.com/v1 and gpt-4o-mini. For a local provider, set AUTOPILOT_LLM_BASE_URL and AUTOPILOT_LLM_MODEL instead; for example, a local OpenAI-compatible service might use http://localhost:11434/v1. Existing process environment variables take precedence, which allows deployment and GitHub Actions settings to override local defaults.

3. Create a test plan. Replace the sample path with your requirements file:

   uv run autopilot plan --requirements examples/requirements/login.md --output artifacts/test-plan.json
   

4. Review the plan, then generate the tests:

   uv run autopilot generate --plan artifacts/test-plan.json --output generated_tests
   

5. Review the generated tests. Update any example URLs and test data for your approved test environment, then run them:

   uv run --directory generated_tests pytest
   

Planning and generation require the configured provider. Review generated tests before running them; generation does not guarantee that they are correct. To rerun a command that would overwrite an existing output, first review the existing files and use its --overwrite option.

For the other available features and development phases, see FEATURES.txt.