Git Mew

GitMew - Your AI kitty for Git chaos

AI-powered commit messages, code reviews, and MR descriptions — all inside VS Code.

Features • Quick Start • Usage • Customization • Providers

Features

Git Mew is your intelligent Git companion that leverages AI to streamline commit creation, code review, and merge history analysis inside VS Code. It helps you move faster without losing visibility into what changed, why it changed, and what still needs attention.

Core Capabilities

Overview
_{A quick tour of Git Mew — sidebar panel, commit graph, and all core features at a glance.}

Generate Commit Message
_{Stage your changes and let Git Mew generate a meaningful, conventional commit message in seconds.}

Review Staged Changes
_{Get AI-powered code review on your staged changes before committing — catch issues early.}

Git Mew now has its own Activity Bar icon. The sidebar gives you a unified workspace for staging, committing, pushing, reviewing, and browsing commit history — all without switching views.

Stage/unstage files individually, by folder, or all at once
Commit with AI-generated messages (sparkle button)
Visual commit graph with branch lines, sync status, and merge detection
Squash local commits with undo support and AI-generated squash messages
One-click push when you have unpushed commits
Merge conflict warnings at a glance
Quick access to all code review workflows and settings

Smart Commit Messages

Automatically generate meaningful, conventional commit messages by analyzing your staged changes. No more staring at blank commit boxes.

Pre-Commit Review

Get AI-powered analysis of your staged changes before committing. Catch potential issues, receive improvement suggestions, and understand the impact of your changes.

Comprehensive Code Review

Generate detailed code reviews for merge requests by comparing any two branches. Includes quality assessment, security considerations, and actionable feedback.

Review Merged Branch History

Inspect code that has already been merged into the current branch by selecting a historical merge commit. This is useful for understanding past work, auditing risky merges, or reviewing work after integration.

Review Selected Commits

Select one or more commits directly from the graph and generate a review from their combined diff. Useful for reviewing a specific set of changes without switching branches or creating merge commits.

Review History

Every review is automatically saved to ~/.gitmew/.histories/ as a Markdown file. Browse, preview, and manage past reviews from the "Histories" sidebar view.

Professional MR Descriptions

Create polished merge request descriptions with smart template selection:

Default: Standard feature/bugfix descriptions
Release: Release branches with changelog links
Hotfix: Urgent fixes with incident tracking

Visual Flow Analysis

Reviews include PlantUML diagrams to visualize code flow, architecture changes, and component interactions. Diagrams can be fixed with AI if syntax errors occur.

Deep Change Explanations

Review outputs now include both a concise summary and a longer Detail Change section to explain the logic, behavior, and implementation impact of code changes in more depth.

Context-Aware Analysis

Git Mew intelligently inspects related files outside the diff to understand runtime flow and catch hidden integration risks.

Security Analysis

A dedicated Security Analyst agent performs OWASP-aligned taint analysis, CWE-tagged vulnerability detection, and auth-flow inspection as part of every review.

Power Features

Multiple LLM Providers: OpenAI (GPT-5.4, GPT-5.4 Mini, GPT-5.4 Nano), Claude (Sonnet 4.6, Haiku 4.5), Gemini (3.1 Pro, 3 Flash), Ollama (local), and custom OpenAI-compatible endpoints
Zero Setup for Ollama: Run AI models locally without API keys
Conventional Commits: Automatic formatting following best practices
Multi-Language Support: Get reviews and descriptions in 8+ languages
Custom Rules: Define project-specific guidelines via .gitmew/ folder
Quick Access: Integrated sidebar in VS Code's Activity Bar and Source Control panel buttons
History-Safe Review Picker: Review Merged Branch shows the 20 most recent merges by default and uses search for older history to avoid UI lag
Review Memory: Cross-session memory de-prioritizes recurring low-value findings and tracks resolution history
Status Bar Menu: Fast access to frequently used commands
Template Publishing: One-click distribution of customization templates
Cancellable Operations: Stop generation mid-flight if needed
Auto-Staging: Automatically stage changes when nothing is staged
Secure Storage: API keys encrypted in VS Code's secret storage

Quick Start

Installation

Install Git Mew from the VS Code Marketplace
Click the Git Mew icon in the Activity Bar to open the sidebar
Open Settings and configure your AI provider:
- OpenAI/Claude/Gemini: Enter your API key
- Ollama: Just select your local model (no API key needed)
- Custom: Provide your OpenAI-compatible endpoint URL and API key
Start staging, committing, reviewing, and pushing — all from the sidebar

Usage

Generate Commit Messages

Never struggle with commit messages again. Git Mew analyzes your changes and creates professional, conventional commit messages.

How to use:

Stage your changes in Git
Click the sparkle icon (✨) in the sidebar or Source Control panel
- Or run: git-mew: Generate Commit Message
Git Mew analyzes your changes and generates a commit message
Review and commit!

Output format:

feat: add user authentication system

Files changed:
- src/auth/login.ts: Implement JWT-based authentication
- src/middleware/auth.ts: Add authentication middleware
- tests/auth.test.ts: Add authentication tests

Implemented secure user authentication using JWT tokens with bcrypt password hashing and session management.

Pro tips:

If nothing is staged, Git Mew offers to stage all changes automatically
Click the pause icon to cancel generation mid-flight
Use the status bar menu for quick access to all commands

Review Staged Changes

Get AI-powered feedback on your changes before committing. Catch issues early and improve code quality.

How to use:

Stage your changes in Git
Click the eye icon (👁️) in the sidebar or Source Control panel
- Or run: git-mew: Review Staged Changes
A dashboard opens with options:
- Select LLM provider and model
- Choose output language
- Add optional task/issue context
Click "Generate Review"
Review the AI analysis, suggestions, and potential issues

What you get:

Code quality assessment
Potential bugs and issues
Improvement suggestions
Summary of changes
Detail Change walkthrough
Security analysis (OWASP-aligned, CWE-tagged)
PlantUML diagrams showing code flow
Related file analysis for context
Observer warnings for hidden risks

Review Merged Branch

Review code that has already been merged into your current branch. This is useful when you want to audit completed work, understand a past merge, or inspect a risky integration after the fact.

How to use:

Check out the branch you want to review from
Click the history icon in the Source Control panel
- Or run: git-mew: Review Merged Branch
A dashboard opens with options:
- Browse the 20 most recent merged branches
- Search older merged branches without loading the entire history into the UI
- Select LLM provider, model, and output language
- Add optional task or issue context
Click "Generate review"
Inspect the result in separate Review and Diff tabs

What you get:

Review generated from the exact merge commit patch
Summary of changes
Detail Change walkthrough
Commit-message context from the merged branch history
Potential bugs, risks, and follow-up suggestions
PlantUML diagrams and repair support when needed

Review Selected Commits

Review a specific set of commits directly from the graph without switching branches or creating merge requests.

How to use:

Open the Graph view in the Git Mew sidebar
Select one or more commits using the checkboxes
Click the "Review" button in the toolbar
A review dashboard opens showing the selected commits
Choose LLM provider, model, and output language
Click "Generate review"

What you get:

Combined diff review across all selected commits
Code quality assessment with security analysis
Detail Change walkthrough
PlantUML diagrams and repair support
Review auto-saved to history for later reference

Generate Code Reviews

Create comprehensive code reviews for merge requests by comparing branches.

How to use:

Commit your changes to your feature branch
Click the merge icon (🔀) in the Source Control panel
- Or run: git-mew: Review Merge
A dashboard opens:
- Select base branch (e.g., main)
- Select compare branch (your feature branch)
- Choose LLM provider, model, and language
- Add optional task/issue context
Click "Generate Review"
Review opens in a new tab with comprehensive analysis

Review includes:

Summary of all changes
Detail Change walkthrough
Code quality assessment
Security analysis (OWASP-aligned, CWE-tagged)
Performance considerations
Improvement suggestions
PlantUML diagrams visualizing architecture
Observer todo list
Related file context for better understanding

Generate MR Descriptions

Create professional merge request descriptions with smart template selection.

How to use:

Follow steps 1-3 from "Generate Code Reviews" above
Click "Generate Description" instead
Description opens in a new tab, ready to copy

Smart templates:

Default: Detected for standard feature/bugfix branches
Release: Auto-selected for release branches (includes changelog)
Hotfix: Auto-selected for hotfix branches (includes incident tracking)

Description includes:

Problem/feature summary
Changes made (grouped by scope)
Related issues and tickets
Testing checklist
Deployment notes (for release/hotfix)

Requirements

Visual Studio Code 1.75.1 or higher
Git installed and configured
API key for your chosen LLM provider (OpenAI, Claude, or Gemini)
- Note: Ollama does not require an API key - just have it running locally

Setup

Install the extension
Run the command git-mew: Setup Model to configure your AI provider
Select your preferred LLM provider (OpenAI, Claude, Gemini, or Ollama)
Enter your API key when prompted (not required for Ollama)
Choose your preferred model

Using Ollama (Local AI)

Ollama allows you to run AI models locally without requiring an API key:

Install Ollama on your machine
Pull a model: ollama pull llama2 (or any other model)
Make sure Ollama is running
Select "Ollama" as your provider in Git Mew
Choose your installed model

Customization

Git Mew is highly customizable. Define project-specific rules, prompts, and behaviors to match your team's workflow.

How Rules Are Loaded (Priority Order)

Git Mew resolves customization files using a three-tier priority chain:

Project (.gitmew/)  >  Global (~/.gitmew/)  >  Built-in defaults

Project-level (.gitmew/ in your repo root): highest priority, overrides everything. Commit to Git to share with your team.
Global (~/.gitmew/ in your home directory): applies to all projects on your machine. Useful for personal preferences.
Built-in defaults: used when no custom file is found at either level.

This means you can set global baseline rules and override specific ones per project.

Getting Started: Publish Rules to .gitmew

The sidebar Settings panel has a "Publish Rules to .gitmew" button. Clicking it opens a picker where you select which template files to copy, then choose the scope:

Project — copies to .gitmew/ in your workspace root (highest priority, team-shareable)
Global — copies to ~/.gitmew/ in your home directory (applies to all projects)

After publishing, open the files and edit them to fit your project. Git Mew picks them up automatically on the next run — no restart needed.

Customization Files

Create a .gitmew/ folder in your repository root (or ~/.gitmew/ for global config) with any of these files:

1. `commit/rules.md`

Define custom rules for commit message generation.

Example:

# Custom Commit Rules

- Always include ticket number in format: [PROJ-123]
- Use present tense for all commit messages
- Maximum subject line length: 72 characters
- Include "Breaking Change:" prefix for breaking changes

2. `review/system-prompt.md`

Completely replace the default review system prompt. Full control over AI behavior.

Use when: You want to fundamentally change how reviews work.

3. `review/code-rules.md`

Add project-specific review rules on top of the default system prompt.

Example:

# Project Review Rules

## Security
- Check all database queries for SQL injection
- Verify authentication on all API endpoints

## Performance
- Flag any N+1 query patterns
- Check for missing database indexes

## Style
- Ensure all functions have JSDoc comments
- Verify consistent error handling patterns

4. `review/agent-rules.md`

Customize internal review agents (flow diagrams, observer checks, security analyst, domain reviewers).

Example:

# Review Agent Rules

## Flow Diagram Agent
- Generate sequence diagrams for authentication flows
- Use class diagrams for data model changes

## Observer Agent
- Focus on security and performance issues
- Flag breaking changes

## Security Analyst Agent
- Prioritize OWASP Top 10 checks
- Flag any hardcoded credentials

5. `description/system-prompt.md`

Customize merge request description generation and template selection.

Use when: You want custom MR description formats or template logic.

Project Structure

your-project/
├── .gitmew/
│   ├── commit/
│   │   └── rules.md
│   ├── review/
│   │   ├── system-prompt.md
│   │   ├── code-rules.md
│   │   └── agent-rules.md
│   └── description/
│       └── system-prompt.md
├── src/
└── ...

Tips

Start with review/code-rules.md — it's additive, so you keep all default review behavior and just layer your project rules on top
Use review/system-prompt.md only when you need full control over how the review agents behave
review/agent-rules.md is useful for steering specific agents (e.g. tell the Security Analyst to focus on your auth layer)
Keep rules concise — the AI reads them as part of the prompt, so shorter and more specific is better
Commit .gitmew/ to Git so your whole team benefits from the same rules
Use ~/.gitmew/ for personal preferences (language, style) that you don't want to impose on the team
To change the commit message language, edit commit/rules.md and update the line **IMPORTANT: You MUST respond in English...** to your target language

Template Variables

You can use dynamic variables inside any .gitmew/ rule file. Git Mew replaces them at runtime before sending to the AI:

Variable	Description	Available in
`{{branch}}`	Current branch name	All rule files
`{{baseBranch}}`	Base branch of the review	Review Merge only
`{{compareBranch}}`	Compare branch of the review	Review Merge only
`{{repoName}}`	Repository folder name	All rule files

Example — review/code-rules.md:

# Project Rules for {{repoName}}

- This review is for branch `{{compareBranch}}` merging into `{{baseBranch}}`
- Treat `main` as the production branch — flag any direct changes to it

Example — commit/rules.md:

# Commit Rules

- Branch: {{branch}} — include ticket number if branch matches `feature/PROJ-*`
- Maximum subject line: 72 characters
- **IMPORTANT: You MUST respond in Vietnamese language. All commit message content must be written in Vietnamese.**

Unknown variables (e.g. a typo like {{branchh}}) are left as-is in the output.

Supported LLM Providers

Git Mew supports multiple AI providers, giving you flexibility and choice.

OpenAI

Models: GPT-5.4, GPT-5.4 Mini, GPT-5.4 Nano

Setup:

Get API key from OpenAI Platform
Select "OpenAI" in Git Mew setup
Enter your API key
Choose your preferred model

Best for: Latest AI capabilities, reasoning tasks, complex code analysis

Anthropic Claude

Models: Claude Sonnet 4.6, Claude Haiku 4.5

Setup:

Get API key from Anthropic Console
Select "Claude" in Git Mew setup
Enter your API key
Choose your preferred model

Best for: Detailed analysis, long context windows, nuanced understanding

Google Gemini

Models: Gemini 3.1 Pro, Gemini 3 Flash

Setup:

Get API key from Google AI Studio
Select "Gemini" in Git Mew setup
Enter your API key
Choose your preferred model

Best for: Fast responses, cost-effective, multimodal capabilities

Ollama (Local AI)

Models: Any model you have installed locally

Setup:

Install Ollama
Pull a model: ollama pull llama2 (or any other)
Start Ollama
Select "Ollama" in Git Mew setup
Choose your installed model

Best for: Privacy, no API costs, offline usage, experimentation

No API key required!

Custom (OpenAI-Compatible)

Models: Any OpenAI-compatible endpoint

Setup:

Select "Custom" in Git Mew setup
Enter your base URL (e.g., https://api.example.com/v1)
Enter your API key
Enter your model name
Configure context window and max tokens

Best for: Self-hosted models, enterprise deployments, custom endpoints

Supported endpoints: Any service implementing OpenAI's /chat/completions API

Privacy & Security

Your data and credentials are protected:

API Keys: Stored in VS Code's encrypted secret storage, never in plain text
Code Privacy: Your code is only sent to the AI provider you choose
Error Tracking: Git Mew uses Sentry for anonymous crash reporting with automatic PII scrubbing. No source code is ever included in error reports.
Local Control: You control which provider and model to use
Ollama Option: Run completely locally with no external API calls

Requirements

VS Code: Version 1.75.1 or higher
Git: Installed and configured
API Key: For your chosen provider (except Ollama)
Internet: Required for cloud providers (OpenAI, Claude, Gemini)

Known Issues

Large diffs (>100KB) may take longer to process
Binary files are detected but not analyzed in detail
Multiple repositories in workspace: only first repo is used
Custom API endpoints must support OpenAI's chat completions format

Roadmap

Future enhancements we're considering:

Streaming responses for real-time feedback
Multi-repository support
Custom commit message templates
Team configuration sharing
Automated testing integration

Have a feature request? Open an issue!

Contributing

We welcome contributions! Here's how you can help:

Report Bugs: Open an issue with details
Suggest Features: Share your ideas in the issues
Submit PRs: Fork, code, and submit pull requests
Share Feedback: Let us know how Git Mew works for you

Repository: github.com/alvin0/git-mew-vscode-extension

License

See LICENSE.md for details.

Author

Châu Lâm Đình Ái (alvin0)

GitHub: @alvin0
Email: chaulamdinhai@gmail.com

Enjoy using Git Mew!

Star us on GitHub | Share with your team | Report issues

Made with love for developers who love clean Git history

alvin0

GitMew - Your AI kitty for Git chaos

Features

Core Capabilities

Sidebar Panel

Smart Commit Messages

Pre-Commit Review

Comprehensive Code Review

Review Merged Branch History

Review Selected Commits

Review History

Professional MR Descriptions

Visual Flow Analysis

Deep Change Explanations

Context-Aware Analysis

Security Analysis

Power Features

Quick Start

Installation

Usage

Generate Commit Messages

Review Staged Changes

Review Merged Branch

Review Selected Commits

Generate Code Reviews

Generate MR Descriptions

Requirements

Setup

Using Ollama (Local AI)

Customization

How Rules Are Loaded (Priority Order)

Getting Started: Publish Rules to .gitmew

Customization Files

1. commit/rules.md

2. review/system-prompt.md

3. review/code-rules.md

4. review/agent-rules.md

5. description/system-prompt.md

Project Structure

Tips

Template Variables

Supported LLM Providers

OpenAI

Anthropic Claude

Google Gemini

Ollama (Local AI)

Custom (OpenAI-Compatible)

Privacy & Security

Requirements

Known Issues

Roadmap

Contributing

License

Author

Enjoy using Git Mew!

1. `commit/rules.md`

2. `review/system-prompt.md`

3. `review/code-rules.md`

4. `review/agent-rules.md`

5. `description/system-prompt.md`