Someone Built Code Review for Agent-Written Docs

I spend a good chunk of my week asking AI agents to write things: specs, plans, documentation. The writing part keeps getting better. The reviewing part has been quietly miserable.

Here's the loop I keep ending up in. I quote a paragraph back into the chat, the agent rewrites half the file, and I can't tell what actually changed. There are no line comments. There is no diff. I'm reviewing a document with the tooling of a text message, and every round trip costs me the one thing review is supposed to protect: knowing exactly what moved.

So a launch I saw this week genuinely caught my eye. DotMD is a native Mac Markdown app built around a single idea: reviewing what your agent wrote should work like a pull request. The agent sends the document over MCP and waits. You select a passage and comment on it, right there in the document. The agent reads your comments and sends fixes back. Each fix lands exactly where your comment was, old text in red, new in green. You accept it, or you send it back.

I haven't used it, so this is a first-impressions note, not a review. But two design decisions in the pitch tell me the maker has lived the same pain I have. First, comments pin to exact passages instead of floating in a chat window — that's the thing chat-based review fundamentally cannot do, and it's the whole ballgame. Second, the MCP server listens on localhost only. Everything stays on your Mac. No account, no cloud. Your specs don't leave the building for the privilege of being reviewed.

The question I'd want answered before paying for Pro: when the agent rewrites a large section, do the comments follow the text, or do they detach? Line-comment mapping under heavy rewrites is the hard problem of every review tool, code or prose. It's the question I left on the launch thread.

Still, the interesting shift here isn't the app itself. It's that "reviewing" is becoming a product category for the agent era. Writing was the easy part. Knowing what changed, and signing off on it — that was always the real work.