I made an MCP that allows your agent to take a screenshot of a webpage. I mainly needed it myself so that I don't have to keep feeding it screenshots of errors while it makes changes to the UI during local development.
I tried it out while building a side project in Cursor with auto-run on and it was able to use it as a feedback loop!
It's open source so any feedback, suggestions, and PRs are more than welcome. Check it out and let me know what you think!