Session logs tell you what your coding agent did. They don’t tell you how it understood the task. Mindwalk’s answer: render your repo as a 3D “code city,” then replay the agent’s session on top of it — searches, reads, and edits flowing through the map like light.
What it actually is
An open-source coding tool: a local Go server plus a React/Three.js frontend. Point it at Claude Code or Codex session logs, and it parses them into file-level touch events mapped onto a deterministically generated 3D layout. Files glow by how deep the agent went — seen, read, or edited — with timeline markers for context compactions and subagent launches. Everything runs locally. Nothing gets uploaded.
Why it’s worth watching
v0.1.0 shipped July 11 and hit the Hacker News front page within a day at 52 points. The author is cosmtrek, maintainer of air — the Go live-reload tool with 23.8K GitHub stars — so this isn’t a weekend toy. The bigger signal: agent observability is becoming its own category. We built tools to run agents. Mindwalk is one of the first that makes an agent’s attention visible.
You Might Also Like
- Openai Codex in Chrome Moves the Coding Agent Into Your Real Browser Session
- Zcode Zhipu z ai glm 5 2 Coding Agent hit hns 1 Spot Chinas Open Answer to Claude Code
- Claude Code Remote Control Just Turned my Phone Into a Coding Terminal and im Weirdly Into it
- 27k Github Stars in Weeks Learn Claude Code by Shareai lab Breaks Down ai Coding Agents Into 12 Lessons
- Agent Kernel Gives any ai Coding Agent Persistent Memory With Just Three Markdown Files
Top AI Product 
Leave a comment