# Zerocheck vs e2e (TesterArmy)

e2e is TesterArmy's open-source TypeScript testing framework, launched on 2 October 2026: agent steps and deterministic assertions in the same test, bring your own model, web and mobile. Zerocheck is an assisted product for teams that do not want to write a test suite: describe the flow or import the release checklist, review the browser result, and save the tests you want to rerun.

## What e2e (TesterArmy) does well

- Fully open source under Apache-2.0, and the first run costs nothing: npx e2e init writes a config and an example test, and npx e2e run passes with no model, no sign-in and no API key. Four days after launch the npm package had about 60,000 weekly downloads.
- One test mixes agent.act, agent.assert and agent.extract with Playwright-style locators and assertions. An agent step that a later assertion verifies is recorded, and later runs replay it without model calls until the app changes, which keeps token spend down.
- Bring your own model with no markup: a ChatGPT, GitHub Copilot, SuperGrok or OpenCode subscription, any API key, a gateway, or a local model. Web on Chromium, Firefox and WebKit; iOS simulators and Android emulators through agent-device; hosted browsers from Kernel and hosted simulators from Expo as optional targets.
- Built for coding agents: one paste prompt sets it up from Claude Code, Codex or Cursor, the package ships a skill, MCP configuration and its docs inside node_modules, and there are migration guides from Playwright, Cypress, Selenium, Detox and Maestro plus a pull-request comment reporter.

## Where e2e (TesterArmy) falls short

- It is a framework, not a service. Your engineers write and maintain .e2e.ts files in TypeScript, pick and pay for a model, and run the infrastructure or hosted targets themselves. TesterArmy's hosted platform is a separate product that uses e2e as its foundation.
- It is pre-1.0. The README says APIs and configuration can still change between minor releases, so a suite written this month may need edits as the framework settles.
- Decide who confirms the expected outcome. agent.assert judges the screen with a model, and only agent steps backed by a later deterministic assertion are cached for replay; a product manager with a release checklist still needs an engineer to turn each item into code.

## How Zerocheck differs

- No test code. Describe a flow in plain English or import the release checklist; Zerocheck drafts each test, runs it in a real browser, and shows the result before you save it. Tests are readable YAML you own, in the web app or in your repository.
- Evidence is the product: step screenshots, console errors, every attempt and the exact test revision, in the web app and in one pull-request comment. A pass after a transient retry stays labeled flaky, and a failed expected outcome is never retried into a pass.
- Runs against your real environments with configured test credentials, locally, in GitHub Actions or in Zerocheck's hosted browser, from the web app, the CLI (also installed as e2e-testing, e2e-qa, qa-tests and other names) or a coding agent over MCP. Hands-on onboarding on a dedicated instance; not open source; Chromium web only; the AI runs through Zerocheck's service.

## Side by side

| Feature | e2e (TesterArmy) | Zerocheck |
| --- | --- | --- |
| Model | Open-source framework; bring your own model and infrastructure | Assisted product: your team describes and reviews, Zerocheck runs and keeps the evidence |
| Who writes the tests | Your engineers, in TypeScript | Your team, in plain English or by importing a checklist; drafts reviewed before saving |
| Test format | .e2e.ts files mixing agent steps and Playwright-style locators | Plain-English YAML, editable and downloadable |
| First run | npx e2e init, then npx e2e run; no account or model needed | Web app or CLI after an onboarding call provisions your instance |
| AI model | Your subscription, API key, gateway or local model; no markup | Zerocheck's service, included in the run allowance |
| Platforms | Chromium, Firefox, WebKit; iOS simulators and Android emulators | Chromium web browsers |
| Evidence | Failure pages under .e2e with the failing step, screenshot, trace and logs | Step screenshots, console errors, every attempt and revision; one updated PR comment |
| Pricing | Free (Apache-2.0); model and hosted targets billed by their providers | Free tier with onboarding; US$500/month Pilot on a dedicated instance |

## Frequently asked questions

### Is e2e by TesterArmy free?

Yes. The framework is open source under Apache-2.0 and the first run needs no model, sign-in or API key. Agent steps use a model you supply: a ChatGPT, GitHub Copilot, SuperGrok or OpenCode subscription, an API key or a local model, billed by that provider, and hosted browsers or simulators from Kernel or Expo are billed by them. Zerocheck has a free tier with an onboarding call, then US$500 per month for one project on a dedicated instance.

### Do I need to write TypeScript to use e2e?

Yes. Tests are .e2e.ts files that import test and expect from e2e, call agent.act and agent.assert for the agentic steps, and use Playwright-style locators and assertions for the deterministic ones. Zerocheck tests are plain-English YAML drafted from a description or an imported checklist and reviewed in the web app before saving.

### Can a team use both?

Yes, for different jobs. Engineers who want a code-first suite with their own model can keep it in e2e. The release checklist the rest of the team checks by hand before a release can run in Zerocheck with evidence in the pull request. Zerocheck does not run e2e tests and does not export them.

Source: [tester-army/e2e on GitHub](https://github.com/tester-army/e2e), [e2e on npm](https://www.npmjs.com/package/e2e), [e2e quickstart](https://e2e.tester.army/docs/quickstart), [Launch post](https://tester.army/blog/introducing-e2e)

---
Canonical: https://tryzerocheck.com/compare/e2e/
Updated: 2026-10-07