Has anyone integrated browser-use-mcp-server with their AI workflow?

I'm trying to automate some research tasks by having Claude browse websites directly through Cursor, and I found browser-use-mcp-server but the docs are pretty sparse. Has anyone here actually set this up? I'm curious about real-world performance—does it handle JavaScript-heavy sites okay, and what's the latency like for complex navigation tasks? Would love to hear what you got working.