Kata: An Intelligent Code Review Tool for Jujutsu Version Control

Kata is an open-source code review tool developed by martint, specifically designed for the Jujutsu (jj) version control workflow. It fills the gap in the jj ecosystem for a code review tool. Its core features include revset-based review definitions, patchset-anchored comments, and a web interface, HTTP API, and MCP interface to support rapid iterative development by AI agents. Whether you are an existing jj user or a technical leader evaluating next-generation version control systems, Kata is worth your attention.

Git, as a standard tool for modern development, has issues such as a steep learning curve, complex history rewriting, and cumbersome branch management. Jujutsu (jj), as a next-generation version control system, separates the user interface from the storage backend and supports flexible reorganization of change sets. However, its ecosystem is not yet complete—especially the lack of a code review tool adapted to the jj workflow—which has become a barrier for teams to adopt jj.

The name Kata comes from the martial arts term 'form', embodying the design philosophy of structured, repeated improvement. Core innovations include:

1. Revset-based review definition: Using jj's revset query language to precisely define the scope of reviews (e.g., specific changes or dependencies), going beyond the limitations of Git branch comparisons;
2. Patchset-anchored comments: Comments are bound to the abstract of code changes (patchsets) rather than line numbers. When code is modified or branches move, comments remain associated with the relevant code, avoiding the comment drift problem in traditional reviews.

Kata provides three interactive interfaces:

• Web interface: A clean and intuitive browser experience supporting side-by-side comparisons, inline comments, and status tracking;
• HTTP API: A RESTful interface that can be integrated into CI/CD pipelines or automated workflows;
• MCP interface: Follows Anthropic's Model Context Protocol, allowing AI agents to query review status and post comments, facilitating AI-assisted development.

Kata obtains repository status through jj's command-line interface and library APIs, fully understanding jj's advanced concepts (such as change descriptions and conflict markers). Review metadata is stored in an independent database, separate from jj's change storage, ensuring data security. The web interface uses modern frontend technologies and supports real-time updates, facilitating synchronous collaboration among multiple reviewers.

Kata is suitable for the following scenarios:

• Early teams adopting jj: Fills the code review gap and enables smooth migration;
• Highly iterative workflows: The patchset anchoring feature adapts to frequent history rewriting;
• AI-assisted development: The MCP interface supports AI agents to participate in reviews;
• Cross-team collaboration: A unified review platform bridging Git and jj workflows.

Kata is an open-source project, and contributors are welcome to participate. The future roadmap includes: integration with more code hosting platforms, enrichment of review automation rules, and improvement of AI agent collaboration features. As a key tool in the jj ecosystem, Kata will drive the upgrade of version control technology stacks.

Kata not only provides efficient code review capabilities for jj users but also surpasses traditional tools with its patchset anchoring and MCP interface design, exploring human-machine collaborative development models in the AI era. It is recommended that existing jj users and technical leaders evaluating next-generation version control systems pay close attention to Kata.