Record. Replay. Automate.
Type in plain English or capture from a manual session. Your tests, prompts, and generated code stay inside your security perimeter. AI inference runs on your LLM. Element34 never sees the test data.
// Generated by Element34 Studio · Code mode test('checkout flow / saved address', async ({ page }) => { await page.goto('/checkout'); await page.getByLabel('Saved address').check(); await page.getByRole('button', { name: 'Continue to payment' }).click(); await expect(page.getByTestId('order-total')).toHaveText('$142.20'); });
Move from manual click-tests to production-grade code in one session, not one sprint.
Time-to-first-testPMs, analysts, and business owners author first-draft tests. Engineers refine and own them.
Low-code modeTest data, prompts, and generated code stay inside the customer perimeter. Audit-logged authoring.
Behind the firewallAI test authoring is the workflow where a software tester uses an AI model to generate executable test code from a higher-level description (plain English) or a guided manual session, instead of writing locators and assertions by hand. Element34 Studio is the AI test authoring module inside SBOX. Two modes, three frameworks, inference on customer-controlled infrastructure.
Hides the code entirely. Non-technical users get a low-code feel, but engineers cannot refine, version, or own the test.
Low-code for non-technical users and production-grade code engineers can refine, commit, and own.
Captures a click sequence as a brittle script. Breaks the moment a CSS class or DOM id changes.
Observes context and intent, generates semantic locators (accessible names, test IDs), and suggests negative cases automatically.
Sends test code and customer data to a vendor's shared inference endpoint. Security review usually blocks it.
Runs inference on the customer's own LLM provider inside the SBOX deployment. No data leaves the customer perimeter.
Writes generic code in isolation. No knowledge of your application, your DOM, or your session state. Output rarely runs as-is.
Generates code tied to the live application, using real DOM, intent, and session state. Commits cleanly to your Git.
Two enterprise QA realities, one workspace:
Persona: product manager, business analyst, business owner.
Type the test in plain English. Studio runs it, records the session, shows the result. Engineers review and own the final test.
Persona: QA engineer moving toward automation.
Run a manual session. Studio observes context, suggests locators and assertions, generates Selenium, Playwright, or Appium code that commits to your repo.
No new framework to learn, no migration. Studio outputs to the standard libraries your team already runs. Commit to your Git, run on your CI, execute on your SBOX grid.
The framework most enterprise SBOX customers already run. Studio generates Selenium in Java, JavaScript, Python, and C#/.NET, dropping into your existing project.
// Studio output, Selenium 4 + Java driver.findElement(By.cssSelector("[data-test='continue']")).click(); wait.until(ExpectedConditions.titleContains("Payment"));
Code mode's default in the latest release. Studio generates Playwright in TypeScript and JavaScript with role-based and accessible-name locators that survive UI redesigns better than CSS selectors.
// Studio output, Playwright + TS await page.getByRole('button', { name: 'Continue' }).click(); await expect(page).toHaveTitle(/Payment/);
Appium tests for native iOS and Android, integrated with SBOX Real Device Cloud. Same authoring workflow, output runs on real devices behind your firewall.
// Studio output, Appium iOS driver.findElement(AppiumBy.accessibilityId("continue_btn")).click(); wait.until(EC.visibilityOf("//XCUI[@name='Payment']"));
AI inference runs on your LLM, inside your deployment, with your keys. Element34 never touches the prompts or the generated code. Your security review reads the architecture, not the disclaimer.
Azure OpenAI, AWS Bedrock, GCP Vertex AI, OpenAI direct, Anthropic direct, or self-hosted open-weights models. Customer picks the provider. Studio routes through it.
The LLM provider key is held by the customer in customer KMS. Element34 never sees a customer LLM key. No vendor master key, no shared inference pool.
Every Studio session, every prompt, every generated code commit is audit-logged to the customer log sink. No black-box AI activity.
Customer IdP via SSO and SCIM. Role-based access for authoring and commit. Customer-defined retention for session recordings and generated code.
Studio authors the test. Auto Heal keeps locators alive when the UI shifts. Automated RCA explains failures. Pulse Report scores quality across runs. Real Device Cloud executes on iOS and Android. All five run inside your private SBOX perimeter.
Plain English to production tests. Two modes for two personas.
You are hereResilient locators that survive UI drift. Cuts maintenance dramatically.
Explore Auto HealCluster failures, summarize root cause, propose fixes. Stop chasing flakes.
Explore Automated RCAAI-driven site health with prioritized opportunities, severity and impact.
Explore Pulse ReportReal iOS and Android devices. Element34-managed, customer-provided, or fully isolated.
Explore Real Device CloudShow us your application, your test framework, your LLM. We demo Studio in your real flow and scope an annual SBOX license.
