Share a rich HTML doc, gather everyone's comments on it, let AI converge and facilitate continuously
I shared htmldrop earlier — a Claude Code skill + CLI that turns any HTML file into a one-command shareable link. But sharing is only half the problem. The other half is discussion — and today that goes to die in buried chat threads, or worse, in competing V1/V2/V3 documents that never actually respond to each other. Someone sends a spec; you reply with your own AI-polished counter-spec; now there are two confident docs talking past each other.
This is the other half — making the shared document the place the conversation actually happens, and then making it easy to act on that conversation.
A link and a password is all it takes to get in — and the page lands as a place to talk, not just read. From there, anyone you've shared it with reacts right on the document: highlight a sentence to comment, select a passage, or drag a box around a diagram — every note pinned to the exact part it's about, no account needed.
Then the part I care about most. Getting comments is easy; knowing what to do with them isn't — some are clean fixes, some are real forks in the road. So you bring the whole thread back to your coding agent: it folds in the clear wins (a vague metric becomes a concrete target) and, on the genuine disagreements, lays out both sides and leaves the call to you. Re-publish and the same link shows the next version, comments intact — the document evolves from the conversation instead of forking into V1/V2/V3 copies that never answer each other.
Example of how it looks like after convergence:
It's open source, works with whatever coding agent you already use (Claude Code, Codex, Gemini CLI, Antigravity…), and your-own-data — self-host the comment backend, or just keep every comment in your repo as the record.
If you want to try it yourself, here's the detailed walkthrough: