🧠 DevBrain AI

The AI-powered developer terminal. No dashboards. No subscriptions. Just your terminal.

DevBrain AI is a zero-friction CLI toolbox that brings the intelligence of frontier LLMs directly into your Git workflow. No browser tabs. No copy-pasting. No context switching.

It automates the repetitive, diagnoses the painful, and reviews the risky — all from a single command.

✨ Features

| Command | What it does | Status | | :--- | :--- | :---: | | brain commit | Analyzes your staged diff and generates a Conventional Commit message | 🟢 Stable | | brain doctor | Diagnoses shell errors and proposes executable fixes in a recursive loop | 🟢 Stable | | brain review | AI code review with severity triage on staged changes, branches, or files | 🟢 Stable |

🎬 Demo

brain commit — Generate commit messages from your staged diff

brain review — AI code review with severity triage

brain doctor — Diagnose and fix terminal errors

🧠 Model-Agnostic by Design

DevBrain is not locked to any single AI provider or model. It routes all requests through OpenRouter, giving you access to every major LLM — free or paid — with a single API key.

# Use a powerful free model
brain config --model "qwen/qwen3-8b"

# Switch to GPT-4o for critical reviews
brain config --model "openai/gpt-4o"

# Use DeepSeek for coding tasks
brain config --model "deepseek/deepseek-r1:free"

Supported models include: GPT-4o, Claude 3.5 Sonnet, Qwen 3, DeepSeek R1, Llama 3, Gemma 3, and every other model on OpenRouter.

🚀 Getting Started

Prerequisites

• Node.js 20+
• An OpenRouter API key (free tier available)

Installation

npm install -g devbrain-ai

Configuration

Create a .env file in your project root:

OPENAI_API_KEY=sk-or-v1-xxxxxxxxxxxxxxxxxxxx
OPENAI_BASE_URL=https://openrouter.ai/api/v1
DEVBRAIN_PROVIDER=openai

Set your preferred model:

brain config --model "qwen/qwen3-8b"

⚠️ Never commit your .env file. It is already in .gitignore.

⚡ Quick Usage

# Generate a commit message from staged changes
git add .
brain commit

# Diagnose a failed command
brain doctor "npm install --legacy-peer-deps"

# Review staged changes before committing
brain review

# Review a specific file
brain review src/auth/middleware.ts

# Review against a branch with security focus
brain review --branch main --focus security

📐 Architecture

src/
├── commands/        # CLI command handlers (commit, doctor, review)
├── services/        # Prompt engineering and AI orchestration
├── providers/       # LLM provider adapters (OpenRouter/OpenAI-compatible)
├── utils/           # Git utilities, logger, spinner
└── index.ts         # CLI entry point (Commander.js)

The provider layer is intentionally thin. Adding support for a new LLM provider requires implementing a single chat() interface — nothing else changes.

📖 Documentation

• brain commit — Conventional Commits generation
• brain doctor — Recursive shell error diagnosis
• brain review — AI-powered code review

🗺 Roadmap

• [ ] brain review --interactive — Apply AI suggestions one by one
• [ ] brain readme — Generate README from codebase
• [ ] brain test — Generate unit tests from implementation
• [ ] Multi-provider support (native Anthropic, Google Gemini)

📄 License

MIT © 2026 Xiaozhi (Steven) Xing