Revset helps you review large amounts of code quickly. It shows your pull request as a dependency graph of the functions, classes and calls that changed. You see what your agent actually did before opening any code, and the problems become easier to spot: a deleted method that's still called somewhere, a helper nothing uses, a signature change with six callers across four files. Connect your GitHub repos and every push is re-analyzed automatically.

Revset shows your pull request as a dependency graph of the functions, classes and calls that changed. You see what your agent actually did before opening any code, and the problems become easier to spot: a deleted method that's still called somewhere, a helper nothing uses, a signature change with six callers across four files. Connect your GitHub repos and every push is re-analyzed automatically.

Hey Product Hunt 👋 I’m Chris, founder of Revset.

Coding agents have moved the bottleneck. Writing code is cheap now, and reviewing it has become the hard part. With agents writing thousands of lines per PR, reading every line has become impossible.

Revset shows you the shape of a PR before the code. Connect your GitHub repos and open a pull request:

🗂 Every changed file becomes a card of the functions, classes and methods it touched, each marked added, modified, renamed or moved, with line counts.
🔗 Edges show what calls and depends on what, and which connections the PR added or removed.
🔍 Click any function to open just its diff, right on the canvas.
🧹 Renames show up as renames, formatting-only changes are labeled, and you can hide tests and unchanged code.
🔄 Every push is re-analyzed automatically, and access follows your GitHub permissions.

Revset doesn’t replace reading code. It tells you where to read, so you can spend your attention on the three functions that matter instead of the forty that don’t.

You can try it today for free on any public GitHub PR: change github.com to revset.dev in the PR’s URL, click Analyze, and the graph is usually ready in under a minute.

Try it on the gnarliest agent-written PR you have and tell me what was confusing, what was missing, and what you wish the graph showed. I’ll be here all day 🙏