gcm-ai

AI-powered Git commit message generator following Conventional Commits.

Reads your staged changes, sends the diff to an LLM, and generates a clean commit message. You accept, edit, or reject before anything is committed.

Install

npm install -g gcm-ai

Requires Node.js 18+.

Usage

git add .
gcm

First run triggers a one-time setup wizard. After that, just run gcm in any git repository.

$ gcm

  Analyzing staged changes...

  feat(auth): add JWT refresh token rotation

  - Implement automatic token refresh before expiration
  - Add retry logic for failed refresh attempts

  ? [Y] Accept  [e] Edit  [r] Regenerate  [n] Cancel

Setup

Runs automatically on first use, or manually with:

gcm --setup

The wizard asks for:

• LLM provider (OpenAI or Anthropic)
• API key (stored locally, never sent anywhere except your chosen provider)
• Preferred model
• Commit message language

Configuration is saved to ~/.gcm/config.json with restricted file permissions.

Commands

| Command | Description | |--------------------|--------------------------------------------| | gcm | Generate commit message from staged changes| | gcm --setup | Run or re-run the setup wizard | | gcm --dry-run | Show message without committing | | gcm --verbose | Print diff sent to LLM and raw response | | gcm --provider X | Override provider for this run | | gcm --model X | Override model for this run | | gcm --help | Show usage information | | gcm --version | Show version |

Supported Providers

| Provider | Default Model | Get API Key | |-----------|-------------------------------|------------------------------| | OpenAI | gpt-4o-mini | https://platform.openai.com | | Anthropic | claude-sonnet-4-20250514 | https://console.anthropic.com|

How It Works

1. Runs git diff --staged in the current repository
2. Parses changed files to infer a commit scope
3. Sends the diff to your configured LLM with an optimized prompt
4. Presents the generated message for your approval
5. Commits only when you explicitly accept

Privacy

• Your API key is stored locally in ~/.gcm/config.json with file permissions 0600
• No telemetry, no analytics, no tracking
• The only network request is to your chosen LLM provider
• Diff content is sent to the LLM and not stored or cached anywhere

Development

git clone https://github.com/samcab28/GCM-AI.git
cd gcm-ai
npm install
npm run build
npm test
npm link    # Test the CLI locally

License

MIT