Urgent.News

What's breaking now, across thousands of outlets.

AI

Meet Clúsia: PR reviews for humans, with your own AI at your side

Most of my code reviews used to happen in a browser tab I'd squeeze in between meetings. Lately a lot of teams hand that job to a bot, and the bot leaves a pile of comments that mostly get resolved without being read. I wanted somewhere calmer to read a change, with an AI I picked sitting next to me. It answers my questions and points at things, and I decide what goes on the PR. That's Clúsia.…

Meet Clúsia, a PR review tool that brings AI assistance to the code review process. Normally, code reviews are done manually, often squeezed between meetings. However, many teams now delegate this task to bots that generate numerous comments, which are typically skimmed over. Clúsia offers a more controlled and personalized approach to code reviews.

Clúsia is a free macOS app, open-source under Apache-2.0, written in Rust. It allows users to review pull requests locally and collaborate with an AI agent named Clúsia. When a reviewer opens a pull request, Clúsia checks it out in a local worktree and enables commenting on specific lines or ranges, as well as replying to threads. All these interactions are stored in a draft on the user's machine. Once the reviewer presses "Finalize review," the draft is submitted to GitHub as a single review.

One of Clúsia's key features is its ability to adapt to changes in the codebase. If the pull request author makes new commits while the reviewer is working on the review, Clúsia moves the existing comments to the new locations and suggests actions such as keeping, moving, marking obsolete, or rewording the comments. Users can then review these suggestions and accept or reject them as needed.

Clúsia communicates with the user's preferred AI model, such as Claude Code or Codex, using a secure and private connection. It never accesses the user's GitHub token or interacts with the Git repository directly. Instead, all actions are performed by the user through a modal window, which presents commands, reasons, and a countdown. This ensures that the user maintains control over the review process.

The app also includes security checks, which analyze the pull request for potential vulnerabilities. These findings include severity levels (high, medium, or low), line numbers, and suggested code improvements. Users can choose to add the findings to their draft, edit them, or dismiss them altogether. If no security issues are found, Clúsia will inform the user accordingly, without claiming the change is entirely safe.

In addition to security checks, Clúsia offers an "Audits" tab, which examines the code for correctness, concurrency, error handling, performance, tests, and documentation. Each aspect is evaluated separately, and the results are displayed in individual tabs. Users can customize the number of audits running simultaneously and add their own custom audits with specific instructions.

Clúsia's visualization capabilities add another layer of insight to code reviews. The "Diagrams" tab generates up to eight diagrams based on the changes introduced in the pull request. These diagrams include flow, dependency, sequence, entity-relationship, use cases, data flow, state machine, class, and a unique "Where the words live" diagram. Users can interact with the diagrams, explore the highlighted code, and comment on changes directly from the visual representations.

To help users stay organized and informed, Clúsia includes a menu bar popover displaying pending PRs, saved reviews that have become stale, and a heatmap of review activity. Users receive notifications for review requests, new commits, replies, mentions, and failed CI runs. They can also set quiet hours to minimize distractions during specific time periods.

Additionally, Clúsia can be controlled via a command-line interface (CLI) using simple commands like `clusia prs`, `clusia open owner/repo#123`, and `clusia review publish`.

Setting up Clúsia is straightforward for users with a compatible Mac. The installation script can be run using Homebrew or Rust, and it requires no administrative privileges. The first build may take around five minutes to complete.

Clúsia is an evolving tool, and the developers welcome feedback from users who regularly review code. To try Clúsia, download and run the installation script from the provided GitHub repository (github.com/rzorzal/clusia). The developers encourage users to provide feedback through issues, stars, or other means to help shape future improvements to the tool.

Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

Read the original at dev.to →

More in AI

More from Saturday 17 October →