I wrote a Cursor<>Figma MCP server to make it easier for AI copilots to implement designs.
It's not perfect, but the resulting HTML and CSS is a lot cleaner than giving AI a screenshot and having it guess.
Link in the replies, take it for a spin and let me know what you think!
@mattparlmer@kalomaze Google is working in the same direction—definitely interested in the idea of paying attention to the entire internet per Jeff Dean.
Probably a combo of larger context windows and other innovations
@theo Everything is customizable and accessible in the CLI via keyboard. Using tmux to flit around projects and sessions. It feels closest to the code.
@mattpocockuk@trq212 Nobody in the replies seems to understand you're using HTML to visualize and approve the changes—for human consumption. Lol.
Awesome idea, I'm immediately adding it to my workflow skills 😄
Just made a little Claude skill called "sidequest" that opens a new interactive subagent in a tmux pane with some context on whatever I'm side questing. So awesome.
@threepointone Parallel subagents to review agent and subagent actual-work to create bespoke visualizations of data flow, architecture, whatever seems relevant.
I dunno—tough problem. Seems very domain-specific (i.e. different visualization for a coding agent vs. analytics researcher)