🤖 Stop Writing Boring Commit Messages. Let a Local AI Do It for You.

How many times have you typed git commit -m "fix stuff"

or git commit -m "wip"

because you just wanted to push your code and grab a coffee?

Writing great, meaningful commit messages is a hallmark of a good developer. It helps your team, makes code reviews easier, and keeps your project history clean. But let’s be honest: when you’re deep in the zone, writing a perfect Conventional Commit is the last thing you want to spend brainpower on.

Enter komit — an AI-powered git commit message generator that runs 100% locally on your machine. No API keys, no internet required, and completely private.

komit

? There are plenty of AI commit generators out there, but most of them send your proprietary diffs to external APIs (like OpenAI or Anthropic). If you are working on a closed-source enterprise project, that’s a massive security no-no.

komit

bridges the gap by leveraging Ollama to run lightweight, powerful open-source LLMs right on your laptop.

feat/auth-login

) to extract the exact Conventional Commit type (feat

) and scope (auth

) before the AI even starts writing!conventional

, simple

, or detailed

commit formats depending on your team's workflow.$EDITOR

) to tweak it before finalizing.When you run komit

, you get a beautiful, clean terminal interface:

┬──────────────────────── Staged files (3) ────────────────────────┬
│  • src/auth.py                                                   │
│  • tests/test_auth.py                                            │
│  • README.md                                                     │
┴──────────────────────────────────────────────────────────────────┴
Branch name: feature/auth-tokens
Model: qwen2.5:7b · Style: conventional

⠋ Generating commit message...

┬──────────────────── Suggested commit message ────────────────────┬
│ feat(auth): add JWT authentication token engine [feature/auth-t… │
┴──────────────────────────────────────────────────────────────────┴

» Choose an action: (y)es, (n)o, (e)dit, (r)egenerate [y]:

y

(Yes) $\rightarrow$ Instantly runs git commit -m "..."

for you.n

(No) $\rightarrow$ Aborts the mission without touching your code.e

(Edit) $\rightarrow$ Launches your environment editor (git commit -m ... -e

) so you can perfect the AI's draft.r

(Regenerate) $\rightarrow$ Feeds the diff back to Ollama for a fresh perspective.You just need Ollama installed and a local model pulled down. We recommend any of these depending on your system specs:

ollama pull qwen2.5:7b

ollama pull mistral:7b

ollama pull llama3.2:3b

You can install komit

using your preferred method:

Via pipx (Recommended for Python CLI tools):

pipx install komit

Via Shell Script (Linux/macOS):

curl -fsSL https://raw.githubusercontent.com/glemiu6/komit/master/scripts/install.sh | bash

Via PowerShell (Windows):

irm https://raw.githubusercontent.com/glemiu6/komit/master/scripts/install.ps1 | iex

Make your life even easier by aliasing it directly in git:

git config --global alias.ai '!komit'

Now, your workflow becomes:

git add .
git ai

By running komit init

, the tool sets up a global configuration file (~/.config/komit/config.toml

) where you can tweak your default model, timeout limits, and preferred output style.

Here is what the different styles look like:

feat(auth): add user authentication [feature/login]

Add user authentication [feature/login]

feat(auth): add user authentication [feature/login]

- Add JWT token generation pipeline

- Implement password security layers using bcrypt primitives

- Append rotation verification routes

komit

is fully open-source under the Apache License 2.0. If you want to check out the code, report a bug, or add a feature, feel free to drop by the repository!

Give it a spin, and let me know in the comments: Do you trust AI to write your commit messages, or are you a commit-message purist? 👇

source & further reading

dev.to — original article Anti Refusal LLM Service A repeatable workflow for paper figures so you stop redrawing them every revision Hermes Agent Changed How I Think About AI Agents: From Answer Engines to Skill-Building Systems