Playwright MCP provides two core operation modes: snapshot mode and visual mode. Snapshot mode operates based on the data structure of accessible snapshots, which has the advantages of fast speed and high stability and is suitable for dealing with scenarios with explicit element references; while visual mode uses screenshots and coordinate positioning, which is more suitable for AI models that require visual recognition. These two modes can be freely switched according to the application requirements. Snapshot mode is suitable for routine automation tasks, while visual mode can cope with more complex visual interaction scenarios. Snapshot mode is enabled by default and can be switched to vision mode by adding the -vision parameter.
This answer comes from the articlePlaywright MCP: Browser Automation MCP Service from MicrosoftThe
