agent-gitv1 0.1.0

AI-powered Git CLI using MCP + Gemini to auto-generate commit messages

agent-gitv1 🤖

AI-powered Git CLI — Uses Model Context Protocol (MCP) + Google Gemini to automatically generate professional commit messages from your diff.

---

How It Works

Your Code Changes
      │
      ▼
[MCP Client] ──stdio──► [mcp-server-git]
      │                        │
      │◄── git diff ───────────┘
      │
      ▼
[Gemini AI] ──► Generate commit message
      │
      ▼
[MCP Client] ──► git add . ──► git commit

---

Prerequisites

Tool	Install
Python ≥ 3.10	python.org
uvx (uv tool runner)	pip install uv
mcp-server-git	Auto-fetched by uvx
Gemini API Key	aistudio.google.com

---

Installation

# 1. Clone / navigate to the project
cd "d:\Vijay Projects\Agent_bhai"

# 2. Install in editable mode (creates the `agent` command globally)
pip install -e .

# 3. Set your Gemini API key
set GEMINI_API_KEY=your_gemini_api_key_here     # Windows CMD
$env:GEMINI_API_KEY="your_key"                  # Windows PowerShell
export GEMINI_API_KEY=your_gemini_api_key_here  # Linux / Mac

---

Usage

agent config — Configure LLM Provider (Start Here!)

Run this to configure OpenAI, Google Gemini, or Ollama.

agent config

What you can configure:

• Google Gemini: Uses your API key and a model like gemini-2.0-flash.
• OpenAI (ChatGPT): Uses your API key and a model like gpt-4o-mini.
• Ollama (Local/Network): Provide the base URL (e.g. http://localhost:11434 or http://192.168.1.50:11434). The CLI will automatically fetch your downloaded models and let you choose one!

---

agent commit — Stage + AI commit message + commit

# In any git repo:
agent commit

# Specify a repo path:
agent commit --repo /path/to/repo

# Generate multiple suggestions (pick one or type your own):
agent commit --suggestions 3

# Verbose mode (shows diff preview + available MCP tools):
agent commit --verbose
agent commit -v

agent commit is now history-aware:

• It uses recent commit messages from your repo to align tone/style.
• It includes changed file names in the LLM prompt for better scoped messages.

agent push — Push to remote

agent push
agent push --remote origin --branch main

Help

agent --help
agent config --help
agent commit --help
agent push --help

---

Example Session

🗂  Repository: D:\my-project

🔌 Connecting to mcp-server-git...
✅ MCP session initialized.

📂 Fetching git diff (unstaged changes)...
   Diff captured (1240 chars).

🤖 Generating commit message with Gemini...

💬 Commit Message: feat(auth): add JWT token refresh endpoint

Proceed with git add + commit? [Y/n]: y

📦 Staging all changes (git add .)...
   Files staged.

✍️  Committing...
   [main a3f12bc] feat(auth): add JWT token refresh endpoint

🚀 Done! Changes committed successfully.

---

Environment Variables

Variable	Required	Description
GEMINI_API_KEY	✅ Yes	Your Google Gemini API key

---

Project Structure

Agent_bhai/
├── agent.py          # Main CLI + MCP client logic
├── pyproject.toml    # Packaging + entry point config
└── README.md         # This file

---

License

MIT