Hey, I’m Vignesh, solo dev. I built Handler, a Mac app for reviewing AI-generated code before it becomes one giant diff. It works with Codex and OpenCode. While the agent is generating code, Handler breaks the work into smaller reviewable edits. Each edit comes with a short explanation: what changed, why it changed, and what it touches. If something feels unclear, you can open a side chat on that specific edit and ask follow-ups like “why this approach?”, “what else does this affect?”, or “redo this part,” without derailing the main agent run. The idea is similar to why stacked PRs are easier to review: smaller, understandable units. Handler applies that at AI generation time, before the change turns into one giant diff. I built it because I noticed myself losing the mental model when agents gave me 300–500 line changes. Tests would pass, the diff would look reasonable, and I’d start skimming instead of really understanding what changed. The bottleneck was no longer writing code. It was understanding the code fast enough to stay in control. What Handler does: - Explains each agent edit as it happens - Lets you side-chat on a single edit without de-railing the main agent (fork chat) - Lets you accept/reject changes edit by edit - Runs on existing Codex and OpenCode setups - Uses worktrees so parallel runs don’t collide - Includes a built-in terminal with a Chrome DevTools-style feel, plus a native JSON viewer/builder for inspecting structured output It’s not trying to replace review tools like CodeRabbit or Greptile. Those review after the fact. Handler sits earlier in the workflow, while the code is being generated, so you can keep the mental model instead of reconstructing it later. I’ve been using it daily and can’t go back to merging agent changes blind. Would love feedback from anyone using coding agents seriously.

Handler

Review AI edits like stacked PRs at generation time.

Review AI edits like stacked PRs at generation time.