Record in the browser,
export to Cypress, Playwright or Selenium
Flows captures your browser actions and converts them into production-ready test scripts. Playwright, Cypress, Puppeteer. AI-powered selectors that survive UI changes.
What Flows does today
Record it once, run it whenever you need
Five things you can rely on right now, from the first click to the report.

Record a test by using your site
Click and type through your app in Chrome, and Flows saves it as a test you can run again.
Run a test in the background
Replay a saved test in a background browser and keep the result to look at later.
Run a whole batch at once
Run a set of saved tests in the background and keep every single result.
Results go straight to BugBoard
Each result is filed in BugBoard under the right test run, so nobody copies results by hand.
Take your tests with you
Export any recorded test as a Playwright script (JS or TS), a Cypress spec or a Selenium script.
What's new
Flows updates, early November 2026
Chrome extension 2.3.18 to 2.3.21 and MCP server 1.73.0 to 1.78.0.
Tests more of your real app
- Video calls. In headless runs (MCP server), run a flow on a page with a live call, using fake camera and microphone feeds.
- Scrolling. In headless runs, a real mouse-wheel scroll, and the step passes only if the list moved.
- File uploads. Drop a file on the page while recording and Flows saves an upload step, marked until you give it a file path.
- API checks. In headless runs, a response check reads the response of the step before it.
- Iframes. Waits and checks on a step inside an iframe read that frame.
Safer to run
- Careful self-healing. In headless runs you can switch on self-healing. A broken button or link is retried only when exactly one visible element has the same text. It never touches delete or payment buttons, and your flow changes only after you approve the fix.
- Header secrets. In headless runs, send a saved secret as a request header, only to the host you set.
- Credential consent. An imported flow cannot send your saved credentials without your approval.
- Local data. The MCP server keeps its local data files readable only by you.
Easier to start
- First-run tour. It points at each control in turn.
- Sign-in. When Flows signs in to explore a site, it ignores invisible elements, and a redirect the page starts on its own no longer counts as the form being sent.
Export to Cypress, Playwright or Selenium
Record in the browser, then export to Cypress, Playwright or Selenium, whichever your team already uses. Full Page Object Model support, clean code, ready for CI/CD.
TypeScript or JavaScript with async/await. Full Page Object Model generation.
Modern E2E with built-in assertions and automatic waiting. cy.* commands.
Headless Chrome automation. Direct CDP access for advanced scenarios.
Built for real QA workflows
Not a toy. We use Flows daily on client projects. Every feature exists because we needed it.
AI-Powered Selectors
Intelligent selector generation using data attributes, ARIA roles, and text content. Creates stable locators that survive UI changes without breaking your tests.
Smart Wait Handling
Automatically detects network requests, animations, and loading states. Generates appropriate waits without hardcoded timeouts, a common source of flaky tests.
Built-in Assertions
Auto-generates visibility, text content, and state assertions. Add custom assertions with a single click.
Edit Before Export
Review and modify recorded steps. Add variables, loops, and conditions directly in the UI before exporting.
MCP Server Integration
Connect Flows to AI agents via Model Context Protocol. Let Claude or other LLMs control browser automation directly.
AI builds your tests from scratch
Give Flows a URL and requirements. It explores, designs, creates, runs, and heals tests autonomously.
EXPLORE
Playwright crawls the site, discovers forms, buttons, and navigation
DESIGN
AI architects a test suite from the element catalog
CREATE
Step groups, elements, and flows persisted automatically
EXECUTE
Suite runs with setup and teardown before and after each test
HEAL
Flows proposes each selector fix and an engineer approves it, so suites need less upkeep
Claude IS the designer. Explore-only mode returns the element catalog, and Claude designs the suite directly. No external API key needed.
Connect via MCP, call flows_generate_from_prompt with a URL and requirements. Works with any MCP-compatible editor.
Set a Google, OpenRouter, or OpenAI API key. The AI explores, designs, and creates without human input.
We automated with AI - every test passed, nothing worked. That's why we build tools that keep humans in the loop.- BetterQA Engineering
Generate and debug with AI
Generate full test suites from a URL, or let Claude analyze failures and fix selectors - all via MCP (Model Context Protocol).

We build our own tools, including an MCP server for Flows. Ask us if you want to use it.
Offline Analysis
Debug flows without the extension running. Data syncs to ~/.flows/
AI Error Explanations
Get human-readable analysis of failures: timeouts, stale elements, click interception
Selector Auditing
Score your flows for stability, identify brittle patterns before they break
Auto-Fix Suggestions
Claude finds alternative selectors from your element library
Agentic Testing Flow
Tell the AI what to test
Playwright explores the site
AI designs and creates suite
Run tests, auto-heal failures
Results sync to BugBoard
MCP Tools Available
flows_generate_from_prompt
One-shot: URL + requirements to full test suite
flows_explore
Discover site elements with Playwright
flows_create_from_design
Create suite from AI-designed JSON
flows_analyze
Debug failures, suggest fixes
flows_fix_selector
Apply selector fixes
integration_run_and_report
Run flows and submit to BugBoard
See it working
Flows, recorded live
A screenshot of a 500 error on a real app. BugBoard writes up the bug and designs the test cases, then Flows automates them and runs them against the app. One test catches a real bug. Recorded live, 7:56.
Ready to record?
See how Flows and BugBoard work together. 15 minutes to understand your testing needs.
Want our engineers to build and run the tests for you? See our test automation service →