KushoAI UI Testing TUI

Writing UI tests by hand is slow and tedious. You interact with a page, then translate every click and input into Playwright code, then do it again for the edge cases, the error states, and the boundary conditions. Most teams end up with happy-path coverage at best.

KushoAI UI Testing TUI changes that workflow. You record your user flow once in a real browser, and the tool uses AI to expand that single recording into a full test suite covering multiple input variations, edge cases, and error scenarios. The heavy lifting is automated. You stay in the terminal the entire time.

The core loop:

• Record a user flow using Playwright's browser recorder. No code required.
• Review the generated Playwright script in your terminal editor and make any adjustments.
• Extend the recording with AI: the tool produces multiple test variations, edge cases, and optional custom scenarios based on plain-English instructions you provide.
• Run the full test suite headlessly or with a visible browser, with optional video recording and detailed HTML reports.

Everything runs locally. Your API keys are stored in ~/.kusho-credentials and never sent anywhere beyond your chosen LLM provider.

If you prefer a guided experience over typing commands, run kusho ui to open the interactive menu. Every step in the flow below — credentials, recording, test generation, and running — is available here with arrow-key navigation and no flags to remember.

This is a good starting point if you are new to the tool. Press Esc or Ctrl+C at any prompt to cancel and return to the main menu.

Clone the repository, install dependencies, and link the kusho command globally so it is available from any directory.

After linking, verify the install worked:

Before recording, tell KushoAI which LLM to use for test generation. Run the credentials command and follow the prompts:

You will be asked to pick a provider and paste in your API key. Three providers are supported:

• OpenAI (default model: gpt-4o)
• Anthropic (default model: claude-haiku-4-5-20251001)
• Gemini (default model: gemini-2.5-flash)

You can override the default model during setup. Your credentials are saved to ~/.kusho-credentials locally and used only when calling your chosen provider. Run kusho credentials again at any time to switch providers or update your key.

Launch the recorder with a URL (or without one to enter it interactively). A browser window opens. Perform the actions you want to test — navigate, click, fill forms, submit. When you are done, close the browser.

After you close the browser, the generated Playwright script opens in your terminal editor. Review it, make any corrections, and save to continue. The script is saved to kusho-tests/recordings/.

Want to try it out before recording your own app? Run the built-in demo against Playwright's sample to-do app:

Once you have a recording, the kusho extend command sends it to the LLM and generates an expanded test suite with multiple variations, edge cases, and error scenarios.

After saving the reviewed script, you will be prompted for optional instructions to guide AI generation:

Generated test suites land in kusho-tests/extended-tests/. You can further refine any generated suite using plain-English edit instructions:

Run any extended test suite from kusho-tests/extended-tests/. Tests run headlessly by default. Add --headed to see the browser, and --record to capture screenshots and video.

After the run completes, you get a detailed HTML report with pass/fail results, screenshots, video (if --record was used), performance timing, and error details for any failures.