hunk

hunk is an open-source terminal diff viewer built for agentic coding workflows. Its job is to help developers review code changes before they apply or merge them. That sounds narrow, but it sits at an important control point: AI coding agents can produce large patches quickly, and humans still need a fast way to inspect what changed. The project is best understood as a review-first layer for developers using tools such as Claude Code, Codex, or other agentic coding systems. A coding agent may propose edits, but hunk focuses on the step after generation and before acceptance. It gives terminal-heavy engineers a way to examine diffs without leaving the command-line workflow that many agent sessions already use. OpenTools reviewed the public modem-dev/hunk GitHub repository and the source signal describing it as a review-first terminal diff viewer for agentic coders. That makes it a tool for software development process, not a code generator by itself. The distinction matters. hunk does not replace testing, code review, or version control. It helps make review more deliberate when AI tools are moving quickly. The strongest use case is human-in-the-loop coding. A developer can let an agent draft a fix, then inspect the proposed patch before applying it. Teams adopting agent workflows can use a review-first pattern to reduce accidental changes, hidden regressions, or broad edits that do not match the original task. CLI-heavy engineers can keep that process inside the terminal instead of switching between a chat window, editor, and web review interface. Because hunk is open source, builders can inspect how it presents changes and decide whether it matches their safety expectations. That is valuable for AI coding workflows where trust is earned through visibility. A good diff viewer should make changes easier to understand, not hide them behind a polished interface. Developers should still run tests, check generated code manually, and use normal pull-request review for production changes. Pricing is currently best described as free open-source access. Any cost comes from the AI coding agent used alongside it, not from hunk itself. Before adopting it, test it on a small repository, compare it with your existing git diff or review tool, and decide whether it makes agent-generated changes faster and safer to evaluate. For best results, pair hunk with a clear rule: no agent-generated patch is applied until a human has reviewed the diff. Start with small tasks, reject broad unrelated edits, and keep normal test and pull-request checks in place. The tool improves visibility; it does not remove engineering judgment.