AtlasEngine

I built Atlas (atlasengine.dev): a local-first AI code reviewer that works with Claude Code, Codex, Cursor, and Opencode via an MCP server. I've been a hands-on CTO in game dev for the past 17 years. Over the last year I've watched the shift from devs writing a few hundred lines a day to agents shipping entire features in an hour. PR-stage reviewers like CodeRabbit catch surface issues, but they miss architectural drift, small decisions that compound across commits. IMO the PR is too late a stage for review. So I built Atlas for myself: automated review on every branch, plus a UI to quickly walk through changes when something looks off. How it works: * Plugs into your coding agent via MCP * Reviews each branch (or PR) locally, code never leaves your machine * Surfaces findings continuously, not just at PR time Demo (1.5 min): https://www.youtube.com/watch?v=0J7O556jBW4 It's free to use and runs on top of your existing AI subscription. I've been dogfooding it for 2-3 months. I'd genuinely value feedback on: * Are the findings useful, or just noise? * Does branch-level review actually change how you self-review? * What would block you from putting this in your workflow? Happy to answer anything in the comments.