I just published a brief Medium article to document an interesting use case to share with the Cursor community here leveraging the new MCP composer agent functionality.
I’m personally really hyped about the implications of this, so I felt like I should share.
It’s not… “safe” necessarily. But it’s working really well for me, allowing the Composer Agent much more direct access to getting relevant stateful information at runtime or testing theories out directly in thread.
Would be interested to hear thoughts and feedback. Anyone else doing cool stuff with the new MCP support?
Great article… and weirdly, I didnt know about the MCPs until your article.
I always run into that problem when I am trying to do something fairly complex for my own capabilities whereby I cant fit into context, because so many debug testing logging error handling that it just becomes imposible.
Especially when Claude was constantly changing code that was already baked and you dont know until too many bit flips Token_Terror happens/
Funny I have been doing this all afternoon:
Just red modal MCP docs, and thats interesting similar to my thing - however I wonder if I can use my sqlite capture in that – Ill post the repo after I finish testing it. Also, R1 really likes to try to avoid work. Smart, but lazy.