md-redline - inline review comments for markdown, readable by AI agents
Comments embedded in the .md file itself—no sidecars, agents read them natively.
Terminal-based spec review tool with real-time AI conversation (Python/Textual)
Closes the AI spec review loop without leaving the terminal like Cursor's inline chat.
Developers using Claude Code and AI coding assistants
1. Brainstorm (with obra/superpowers) 2. Create a spec.md 3. Review and iterate on the spec 4. Approve and let the AI implement 5. Code review, follow-up fixes, and ship
Step 3 is where it gets painful. After the AI creates the spec, you need to somehow open the file (in my case, split pane, floating window then neovim), read through it, then switch back to Claude Code to give feedback. It breaks the flow completely. You can ask Claude to walk you through it section by section, but then your feedback isn't precise and you can't go back to a previous section once you move on.
The usual code review tool is round by round and it does not really feel fast and interactive enough for me.
I built `revspec` (Review Specification) to fix this. It's a terminal TUI that lets you review AI-generated specs with real-time AI conversation, like inline code review comments, but for specs. You navigate with vim keybindings, press `c` on any line to comment, and Claude replies right there in the thread. When you're done, press `S` to submit and the AI rewrites the spec with your feedback incorporated.
It comes with a Claude Code plugin so the whole flow is seamless, the AI launches the TUI, watches for your comments, replies in real-time, and rewrites the spec when you submit.
GitHub: https://github.com/icyrainz/revspec-py
Install: `pipx install revspec`
Claude Code plugin: `claude install icyrainz/revspec-py`
It's MIT license. I've been dogfooding this for a while using revspec to review the specs for revspec's own features. Let me know what you think!
Comments embedded in the .md file itself—no sidecars, agents read them natively.
GUI wrapper around git diff with LLM walkthrough, but delta and GitHub already do this.
Finally solves the pain of reviewing 1000-line agent plans without losing context between iterations.
Inline terminal comments beat describing line numbers to your AI agent.
Another doc commenter, but tied to a specific chat-bot ecosystem called Claw.
Structured annotation output pipes directly back to AI agent workflows.