FlowTap

FlowTap came from a specific frustration: tests can pass while the screen still looks broken. It runs a React Native app, walks through flows, captures screenshots, checks them with vision models, and produces an HTML report with concrete UI issues.

Visual QA in mobile is manual, boring, and easy to skip. Traditional tests do not tell you that text overflowed or a layout broke on another device.

A developer should be able to run one tool and get a useful issue list with context: screen, flow location, issue type, and priority.

• CLI for testing React Native flows.
• Supervisor plus discovery, testing, and vision-analysis agents.
• Screenshots, screen graph, and UI issue detection.
• HTML report with journey, issue list, and diffs.
• Run quality evals: coverage, depth, efficiency, and delegation quality.

Agents split the work between exploration, test execution, and image analysis. The vision model reviews UI like a reviewer: spacing, overflow, hierarchy, readability, and accessibility.

TypeScript monorepo, Turborepo, Mastra.ai, Maestro, OpenAI/Gemini/Anthropic models, and an HTML report. Private for now. I will open it when I think it is ready, not earlier.

• Local CLI first, no premature cloud dashboard.
• A report people can read, not only JSON for machines.
• Agents as users of developer tools, not only application code.
• Public version only after the basics are polished.

Working MVP that combines agents, vision, mobile automation, reporting, and evals in one system.

Finish it for a public version, then add GitHub Actions, PR comments, device matrix, and a cloud dashboard.