What is mmit?

mmit is a command line tool that writes git commit messages for you using AI. After you stage your changes with git add, running mmit reads the staged diff, sends it to an AI provider of your choice, OpenAI, Anthropic, Gemini, or OpenRouter, and generates a conventional commit message with a subject line and, when useful, a longer body explaining the change. You can accept the suggestion, edit it, or ask it to try again before anything is committed. Setup starts with mmit init, which walks you through picking a provider and entering an API key, or you can set the relevant key as an environment variable instead. Beyond generating single commits, mmit can also build a changelog from your commit history, grouping entries into sections like Added, Fixed, Changed, and Breaking Changes, and can either print that changelog or write it into a CHANGELOG.md file. A release command goes a step further: it looks at the commits since your last tag, works out whether the change counts as a major, minor, or patch release based on conventional commit markers, updates the changelog, bumps the version number in package.json, commits that as a release commit, and creates a matching git tag. Configuration lives in a project level or global JSON file where you can set your preferred provider, model, and which commit types count during changelog generation. The tool also supports non interactive flags for previewing output without committing or skipping the review prompt entirely, which is useful in scripts. It is released under the MIT license and installed globally through npm.