Managing complex design docs with Claude or Cursor is a UX nightmare. Once you start using LLMs for heavy planning or long-form documentation, the standard chat interface completely breaks down. Conte

I needed a way to structure this feedback loop without the chaos, so I built R3. It’s a lightweight local web UI specifically for organizing the review process of these docs. Instead of digging through endless chat history to find one specific suggestion from an agent, R3 lets me manage the review flow cleanly.

I’ve been running this on my own local projects for about a week. From a productivity standpoint, it’s transformed the review phase from a messy chat session into actual engineering work. It makes providing structured feedback back to my AI agents much more efficient, which is critical for maintaining a clean dev loop when you're working with heavy documentation.

The local setup is still a work in progress—I'm optimizing it to handle larger files faster—but it’s already functional for my daily workflow.

https://github.com/example/r3