mcp will probably be left behind in the future. it was a bad design from the start. anthropic themselves released skills to "fix" the mcp mess. skills are very new but the idea is great. we still are early days but i think it could allow models to use tools more effectively.

we're planning to add an installation step + auth step (which many of the skills require) so that that part get's handled in one single step instead of having to do everything manually

it's basically a cli for controlling a browser. the idea is that an agent like claude code would use it for validating something that it just did like changing something on the UI

What browser? My question comes from security, adding that skills just provides a line of bash, with no further info. I checked the .md file but it just lists a list of commands with agent-browser.

agent-browser is built on top of Playwright. Playwright uses a version of Chromium.

I was looking at this earlier. Has anyone used it? Is it useful compared to the Playwright MCP or Claude's Chrome plugin?

i thought of that but i didn't want to build a job to migrate that to the db. maybe we'll go that route.