ZeroStep
ZeroStep is an AI tool that enhances Playwright tests with the power of GPT3.5 and GPT4. It offers an ai() function that allows users to si...
Last verified:
What is ZeroStep?
ZeroStep is an AI‑powered library that adds natural‑language test steps to Playwright end‑to‑end tests. Instead of writing fragile CSS selectors or XPath locators, you pass plain‑text instructions to ZeroStep’s ai() function, and the backend uses a model (like GPT‑3.5 or GPT‑4) plus an understanding of the current page state to resolve those instructions into concrete Playwright actions, assertions, or queries at runtime. This lets you script complex UI interactions such as filling forms, navigating calendars, searching on Google, clicking in dropdowns, and validating content without tying your tests to specific DOM structures.
The tool is designed to sit directly inside existing Playwright test suites, so you can sprinkle AI‑assisted steps into as few or as many tests as you like without overhauling your workflow. Example use cases include booking a timeslot on a scheduling page, creating a Salesforce opportunity, scraping a stock price from Yahoo Finance, checking revision history on Wikipedia, or validating labels counts on a GitHub repo—all expressed as human‑readable prompts rather than selector‑heavy code. Because the instructions are semantic, the tests can remain more stable when the UI is refactored, as long as the user‑visible behavior stays similar.
ZeroStep targets developers, QA engineers, and SDETs who already use Playwright for browser automation and want to reduce the time spent writing and maintaining brittle selectors. It suits teams that ship UI changes frequently, teams practicing TDD on E2E flows, and anyone who wants to encode product‑level expectations in plain language rather than low‑level DOM details. The system is especially useful when you need to automate intricate or highly dynamic screens, legacy systems with inconsistent markup, or third‑party applications where you cannot control the underlying HTML structure.
ZeroStep pricing
Pricing model: Free
ZeroStep offers a free tier at $0 per month with 500 ai() function calls per month and unlimited users. The Team plan is $20 per month and includes 2,000 ai() function calls per month and unlimited users. Both tiers allow you to integrate the library into your Playwright tests without changing your development workflow.
ZeroStep pros
- Runs directly inside existing Playwright test suites
- Requires zero change to Playwright workflow
- No CSS selectors or XPath locators ever needed
- Uses natural‑language instructions that are easy to read and modify
- AI determines actions at runtime based on screenshots and DOM snapshots
- Less brittle to UI refactoring and layout changes
- Speeds up creation of complex E2E test flows
- Enables TDD‑style E2E testing by describing behavior before UI is finalized
- Can handle dynamic, third‑party, and legacy pages with inconsistent markup
- Supports both actions and assertions in plain text
- Can query page content and return values to test logic
- Works with many real‑world apps such as Calendly, Salesforce, Google, Wikipedia, Yahoo Finance, and GitHub
- Open‑source JavaScript library part of the stack
- Integrates with modern Playwright patterns and utilities
- Free tier available for experimentation and small projects
ZeroStep cons
- Depends on external AI backend and API availability
- Usage limited by ai() function call quota per tier
- Introduces latency while prompts are resolved via WebSocket and model inference
- May occasionally misinterpret instructions or produce unexpected element interactions
- Less transparent than hand‑written selectors for very precise control
- Adds an external dependency and token‑management requirement to the test environment
- Behavior may change if the underlying AI model is updated
- Less suitable for safety‑critical or highly deterministic regression scenarios where predictability is paramount
Frequently asked questions about ZeroStep
What is ZeroStep?
ZeroStep is an AI‑powered library that adds natural‑language test steps to Playwright end‑to‑end tests. You call the ai() function with a plain‑text instruction, and ZeroStep’s backend uses an AI model plus a snapshot of the current page to translate that instruction into concrete Playwright actions, assertions, or queries at runtime, eliminating the need for CSS selectors or XPath locators.
How does ZeroStep integrate with Playwright?
ZeroStep integrates by providing a ai() function that you import into your Playwright test files. You keep using Playwright’s context and page objects while passing natural‑language instructions to ai(); the library communicates with ZeroStep’s backend over a WebSocket to resolve those instructions into Playwright commands that then execute in the browser without changing your overall test structure or workflow.
Do I still need to write any selectors when using ZeroStep?
No selectors are used with ZeroStep; the tool is designed so that you never write CSS selectors or XPath locators. Instead, you describe what you want to do in plain text, such as clicking a button, filling a field, or asserting text, and ZeroStep’s AI assistant determines the appropriate actions based on the current page state.
What kind of tasks can ZeroStep automate?
ZeroStep can automate tasks such as filling forms with realistic values, navigating calendars to book the next available timeslot, searching on Google and verifying the first organic result, creating records in Salesforce, extracting stock prices from finance pages, reading revision history on Wikipedia, and counting labels or issues on GitHub—all expressed as natural‑language prompts to the ai() function.
Is ZeroStep suitable for highly dynamic or third‑party UIs?
Yes, ZeroStep is particularly useful for dynamic, third‑party, or legacy UIs where the DOM structure is inconsistent or out of your control. Because the tests are based on semantic descriptions rather than fixed selectors, they can be more resilient to markup changes as long as the user‑facing behavior remains the same.
How does ZeroStep affect test stability and reliability?
ZeroStep can improve stability by decoupling tests from brittle selectors, making them less likely to break when CSS classes or element hierarchies change. However, since it relies on an AI backend, occasional misinterpretations or latency can introduce new failure modes, so it is best used for exploratory and behavior‑driven scenarios complemented with traditional selectors for critical paths.
What is the role of the ZeroStep backend?
The ZeroStep backend runs a proprietary service that receives your ai() prompts along with a screenshot and DOM snapshot of the current page over a WebSocket. It uses an integration with OpenAI models plus additional logic to map those natural‑language instructions into a sequence of Playwright commands, which are then sent back to your test runner to execute in the browser.
Do I need an API key to use ZeroStep?
Yes, you need a ZeroStep API token to authenticate your tests against the backend. You typically obtain this token after signing up on the ZeroStep website and then store it in an environment variable named ZEROSTEP_TOKEN, which the library automatically reads when running your Playwright tests.
How is ZeroStep priced and what are the usage limits?
ZeroStep offers a Free tier at $0 per month that includes 500 ai() function calls per month and unlimited users. The Team tier is $20 per month and provides 2,000 ai() function calls per month with unlimited users. Both tiers allow you to integrate the library into your Playwright test suites without changing your development workflow.
Can I mix ZeroStep steps with regular Playwright code in the same test?
Yes, you can freely mix ZeroStep ai() calls with regular Playwright code in the same test file. For example, you can use ZeroStep to fill a form or click a button and still use Playwright’s page.waitForURL, page.waitForSelector, or expect statements for assertions and navigation, giving you a hybrid approach that combines AI‑assisted steps with explicit control where needed.