For the past 8 months or so I’ve been working for CodeScene and about 6 of those I’ve been mainly working on the CodeScene CodeHealth MCP. I’m lucky to be working on things for developers that I myself love using and the CodeHealth MCP is something I can’t imagine not having anymore in this new age of software development that is fully agentic.
I haven’t been writing code by hand since December of last year now, but I’m not a vibe coder, and I can’t imagine doing agentic programming without tooling to ensure high quality output, as it will inevitably result in horrible software quality. If anything, with the CodeHealth MCP my code quality is actually better than it was when I wrote code myself.
The most notable feature, to me, is the ability to set up a self-correcting loop where AI will make changes and then run the Code Health Review MCP tool on those changes, improving its own creation with feedback from that review. The review is deterministic and always the same regardless of what AI model you use, so it’s perfect as a compass for AI which is inherently non-deterministic.
This then allows you to focus on what really matters - systems thinking, planning, architecture - instead of cleaning up slop after a thing that is incapable of learning, bringing back joy to software development that otherwise started getting lost for me.
Previous to the CodeHealth MCP I opted to not really use any AI directly as my experience was mostly negative. I didn’t want my job to be that of a glorified slop cleanup specialist, and although I miss the act of writing code in the sense of poetry, I do enjoy the new, wider scope, more holistic software engineering a lot.