⚠️ Affiliate Disclosure: This post is not an affiliate associated review. It may become one in the future and this disclosure will reflect that. This does not affect my reviews and I only recommend tools I genuinely believe in.
diff-tour is a free, MIT-licensed command-line tool that turns a git diff into a guided tour you step through with the arrow keys — built on the premise that if AI now writes the code, the human skill that matters is reading it like a senior engineer. I don’t test dev tools hands-on, so this is an information post, not a review. Everything below comes from the project’s README and its Product Hunt launch, checked on 3 November 2026.

What Is diff-tour and What Does It Do?
Run it in a git repository and it uses the AI coding CLI you already have — Claude Code, Codex CLI or OpenCode 2, with Gemini CLI and Copilot CLI support marked experimental — to build a tour of a change: what it’s for, which files to read in what order, senior-engineer questions like “what if it runs twice?”, and a clear split between must-fix problems and nits. Questions come before the answers, so you can guess first, and you can press ? to ask about any step. It’s read-only, with no server and no telemetry, so it can’t mangle your branch.
Install is a one-line curl script on macOS and Linux, or go install with Go 1.27+ anywhere; Windows is untested so far. It’s an early release from maker Dai Aoki, published under the init-co-jp GitHub organisation, and — unusually — the repo doesn’t accept pull requests, only issues.
Who Is diff-tour For, and What Does It Cost?
Developers reviewing AI-generated changes, juniors learning to read real diffs, and educators. The tool itself is free; it spends your existing AI plan’s allowance. The README’s own estimate is that a tour of roughly 250 changed lines on Claude Code used the equivalent of $1.00–$1.50 at API prices — cheap for a thorough review, but worth knowing before touring a 5,000-line refactor. Reopening a saved tour costs nothing.
Used diff-tour? Tell Me How You Find It
I’m not the right person to judge a code-review tool — you might be. If you’ve used diff-tour on real pull requests, leave a comment. I’d especially like to hear whether the must-fix/nit split matches what a human reviewer would have flagged.

Leave a Reply