AI NewsDev toolsAnnouncement

Reviu is a native desktop app for reading every diff an AI agent writes before any of it reaches a branch

Reviu is a free native desktop app that runs Claude Code, Codex, Gemini and other ACP agents in durable sessions and lets a developer read every diff, add inline comments and rewind turns before anything is committed.

AI News

Editorial2 min read

LinkedInX
Reviu desktop app for reviewing AI-written code

Why it mattersAgent coding without a review step turns code into a stream of pull requests nobody read, and a reviewer that sits where the agent writes keeps the boundary between "model wrote it" and "I checked it" readable.

Agent coding today produces code faster than the person running it can read.

Reviu, launched on Product Hunt on 5 October 2026, is a native desktop app that runs Claude Code, Codex, Gemini and other ACP-compatible agents in durable sessions, and sits on the diff output so the user has to read it before anything moves. The free tier covers agent sessions, local diff review, worktrees, checkpoints, an editor and terminal, and local Git workflows. A Pro tier adds GitHub pull-request review, checks, merge and notifications. Joris Gallot built it; this is his second launch on Product Hunt, with the first on 13 April 2026.

The workflow it imposes

A session in Reviu runs an agent against a worktree. Every turn produces a diff. The user reads the diff for that turn in a review panel beside the code, adds inline comments with file and line context, and either sends the comments back to the agent to keep working or accepts the turn.

Checkpoints are the second part. A turn that goes wrong can be undone by name, which is a different experience from a Git reset: the agent and the file state both roll back together, so the next turn starts from a clean slate rather than from the half-fixed one.

The maker frames the design in one line in the launch: "I trust agents more when the boundary is clear: code the agent wrote vs code I actually reviewed." That is a specific position. Reviu's job is not to make the agent write better code; it is to keep the human reviewer in the loop while the agent is still writing it, instead of after a thousand lines have landed in a pull request.

Where it sits against the alternative

The alternative is the setup most teams run today: the agent writes into the working tree, the developer looks at a git diff output every so often, and the review happens on the merged pull request. Three failure modes come with that. A turn that produces 400 lines of plausible-looking code is read as a block instead of in sequence, so a bug introduced halfway is invisible. There is no way to roll the agent back cleanly, which forces the user to stop and clean the tree by hand. And comments in the pull request happen after the agent has moved on, so the agent cannot act on them.

Reviu is one answer to all three. It is a free native desktop app rather than a VS Code extension, which is a trade: it is one more window to keep open, in exchange for a review surface built for diffs instead of grafted onto an editor. The GitHub pull-request review is behind the Pro paywall; everything else, including the ACP agent runner, is free.

Source

This item was written by an AI system from the linked source. Reveneau is responsible for what it publishes.

Share
LinkedInX