Dev.to
7/20/2026

Playwright can test Chrome extensions. So why does my AI agent still need my help?
Short summary
AI coding agents can write Chrome extension code but can't see the result — humans become the agent's eyes and hands for build, load, screenshot loops. Playwright supports Chrome extensions and Electron, but the setup is complex enough that agents struggle to self-verify. The author built KamoX, a local HTTP API wrapping Playwright and platform-native tooling behind simple endpoints like /check-ui and /rebuild so agents can verify UI without human intervention.
- •AI agents write code but can't verify non-browser UIs (Chrome extensions, Electron, VSCode extensions) without human help
- •Playwright supports these platforms but setup is too complex for agents to self-manage
- •KamoX wraps Playwright and native tooling behind simple HTTP endpoints for agent-driven UI verification
Generated with AI, which can make mistakes.
Is this a good recommendation for you?



