Your git history, ghostwritten by a language model

What it does OpenCommit is a configurable glue layer between git and a wide roster of LLM APIs. It reads your staged changes and generates a commit message—optionally with GitMoji, a longer description, or in another language—using whatever provider you already have a key for. It also integrates with local @commitlint rules so the generated message matches your project’s conventions.

The interesting bit The repository dogfoods itself aggressively: the maintainers claim every commit in the project was authored by the tool. It also treats provider choice as a first-class config dimension rather than an afterthought, caching available models for seven days to avoid redundant API calls.

Key highlights

• Supports OpenAI, Anthropic, Azure, Google Gemini, DeepSeek, Flowise, AIMLAPI, and local Ollama models.
• Runs fully offline via Ollama, with configurable endpoints for remote or Docker-hosted instances.
• Can switch between its own conventional-commit prompt generator or integrate with local @commitlint configuration.
• Generates commit messages in multiple languages; locale is configurable globally or per repository.
• Passes standard git commit flags through transparently.

Caveats

• The OCO_WHY feature, which promises to explain the rationale behind changes, is explicitly marked as work-in-progress and requires a not-yet-built RAG/indexing system.
• The optional “push to remote” prompt is flagged for deprecation.

Verdict Worth a look if you are tired of writing “fix: typo” but refuse to be locked into a single AI provider. Skip it if you already have a shell alias that pipes git diff to curl and you like it that way.