@codemieai/code 0.0.2 → 0.0.3

package/README.md

@@ -1,20 +1,19 @@
-# CodeMie Code
+# CodeMie
 
-> AI coding assistant and CLI wrapper for managing multiple AI coding agents
+> CLI wrapper for managing multiple AI coding agents
 
-CodeMie Code is a unified npm package that provides both a built-in AI coding assistant and a CLI wrapper for managing multiple AI coding agents (Claude Code, Aider, Codex, etc.).
+CodeMie is a unified CLI tool for installing, configuring, and running multiple AI coding agents (Claude Code, Codex, etc.) from a single interface.
 
 ## ✨ Features
 
-- 🤖 **Built-in AI Assistant** - Ready-to-use coding assistant powered by any AI provider
-- 🔧 **CLI Wrapper** - Manage and run multiple AI coding agents from one interface
-- 🛠️ **Complete Toolset** - 15 tools including filesystem operations, git commands, and code execution
-- 🔌 **MCP Support** - Model Context Protocol integration for extensible tool support
-- 🔒 **Security First** - Path validation, symlink protection, and command sanitization
-- 📝 **Diff-based Editing** - Smart file editing with indentation preservation
-- 🌐 **Provider Agnostic** - Works with any AI provider (OpenAI, Anthropic, Azure, etc.)
-- 🎯 **Project-aware** - Operates within allowed directories with ignore patterns
-- 🚀 **Zero Configuration** - Works with environment variables only
+- 🔧 **Unified CLI** - Manage multiple AI coding agents from one interface
+- ⭐ **Built-in Agent** - CodeMie Native ready to use immediately (no installation required)
+- 🤖 **Multi-Agent Support** - Claude Code, Codex, and more
+- 🛠️ **Environment Management** - Centralized configuration for all agents
+- 🚀 **Zero Hassle** - Install and run agents with simple commands
+- 📦 **Provider Agnostic** - Works with any AI provider (OpenAI, Anthropic, Azure, etc.)
+- 🎯 **Task Execution** - Single task mode with `--task` flag for automation
+- 🖼️ **Clipboard Integration** - Automatic image detection from system clipboard
 
 ---
 
@@ -46,9 +45,8 @@ npm link
 ### Verify Installation
 
 ```bash
-# Check if commands are available
+# Check if command is available
 codemie --help
-codemie-code --help
 
 # Run health check
 codemie doctor
@@ -58,860 +56,450 @@ codemie doctor
 
 ## 🚀 Quick Start
 
-### 1. Set Up Environment Variables
+### Installation → Setup → Use
 
-CodeMie Code requires AI provider credentials:
+CodeMie follows a simple three-step workflow:
 
-```bash
-export CODEMIE_BASE_URL="https://your-ai-proxy.com"
-export CODEMIE_AUTH_TOKEN="your-auth-token"
-export CODEMIE_MODEL="claude-4-5-sonnet"
-```
-
-**Add to your shell profile** for persistence:
-
-```bash
-# Add to ~/.bashrc or ~/.zshrc
-echo 'export CODEMIE_BASE_URL="https://your-ai-proxy.com"' >> ~/.bashrc
-echo 'export CODEMIE_AUTH_TOKEN="your-auth-token"' >> ~/.bashrc
-echo 'export CODEMIE_MODEL="claude-4-5-sonnet"' >> ~/.bashrc
-
-# Reload shell
-source ~/.bashrc
-```
-
-### 2. Test Connection
-
-```bash
-codemie-code test
 ```
-
-**Expected output:**
-```
-✓ Configuration loaded
-✓ Testing connection...
-✓ Connection successful
-✓ Model: claude-4-5-sonnet
+1. Install → 2. Setup (Wizard OR Manual) → 3. Use (Install & Run Agents)
 ```
 
-### 3. Start the Assistant
+### Step 1: Install
 
 ```bash
-# In current directory
-codemie-code
-
-# In specific directory
-codemie-code /path/to/your/project
-
-# Execute one-off task
-codemie-code exec "Add error handling to api.ts"
-```
-
-### 4. Interact with the Assistant
-
-```
-╔═══════════════════════════════════════╗
-║       CodeMie Code Assistant          ║
-╚═══════════════════════════════════════╝
-
-Working directory: /your/project
-Model: claude-4-5-sonnet
-
-Type "exit" to quit, "clear" to clear history
-
-You: List all TypeScript files in this project
-You: Show git status and recent commits
-You: Create a new file called utils.ts with helper functions
+npm install -g @codemieai/code
 ```
 
----
-
-## 📚 Configuration
+### Step 2: Setup (Choose One Method)
 
-### Required Environment Variables
+#### Method A: Interactive Setup Wizard (Recommended)
 
-| Variable | Description | Example |
-|----------|-------------|---------|
-| `CODEMIE_BASE_URL` | AI provider base URL | `https://your-ai-proxy.com` |
-| `CODEMIE_AUTH_TOKEN` | Authentication token | `your-token-here` |
-| `CODEMIE_MODEL` | Model to use | `claude-4-5-sonnet` |
-
-### Optional Configuration
-
-#### Runtime Configuration
+Best for most users - guided configuration with connection testing:
 
 ```bash
-# Enable debug logging
-export CODEMIE_DEBUG="true"
-
-# Timeout for AI requests (seconds)
-export CODEMIE_TIMEOUT="300"
-
-# Load specific MCP servers
-export CODEMIE_MCP_SERVERS="time,context7"
+codemie setup
 ```
 
-#### MCP Configuration
+**What it does:**
+- ✅ Guides you through provider selection
+- ✅ Prompts for credentials
+- ✅ **Validates credentials** via `/health` endpoint
+- ✅ **Fetches available models** via `/v1/models` endpoint
+- ✅ Shows real-time model list (no hardcoded options)
+- ✅ Saves to `~/.codemie/config.json`
 
-Configure MCP (Model Context Protocol) servers in `~/.codemie/config.json`:
+**Supported Providers:**
+- AI/Run CodeMie (Unified gateway)
+- AWS Bedrock (Claude via AWS)
+- Anthropic (Direct API)
+- Azure OpenAI (for GPT models and Codex)
+- Custom LiteLLM Proxy
 
-```json
-{
-  "mcpServers": {
-    "context7": {
-      "command": "npx",
-      "args": ["-y", "@upstash/context7-mcp"]
-    },
-    "time": {
-      "command": "uvx",
-      "args": ["mcp-server-time"]
-    }
-  }
-}
-```
+#### Method B: Manual Configuration Guide
 
-Or use environment variable to specify which servers to load:
+Best for automated setups or when you have credentials ready:
 
 ```bash
-# Load specific MCP servers
-export CODEMIE_MCP_SERVERS="context7,time"
-codemie-code
-
-# Or with CLI flag
-codemie-code --mcp-servers context7,time
+# View detailed setup guide with examples
+codemie env
 ```
 
-When no `--mcp-servers` flag is provided, all servers configured in `~/.codemie/config.json` are loaded automatically.
+### Step 3: Use Agents
 
-### Configuration Methods
+#### Try the Built-in Agent (Recommended for New Users)
 
-#### Method 1: Shell Environment (Recommended)
+**CodeMie Native** is built-in and ready to use immediately:
 
 ```bash
-# Add to ~/.bashrc or ~/.zshrc
-export CODEMIE_BASE_URL="https://your-ai-proxy.com"
-export CODEMIE_AUTH_TOKEN="your-token"
-export CODEMIE_MODEL="claude-4-5-sonnet"
-
-# Reload shell
-source ~/.bashrc  # or ~/.zshrc
-```
-
-#### Method 2: Project .env File
+# Start interactive session
+codemie run codemie-code
 
-```bash
-# Create .env in your project
-cat > .env << EOF
-CODEMIE_BASE_URL=https://your-ai-proxy.com
-CODEMIE_AUTH_TOKEN=your-token
-CODEMIE_MODEL=claude-4-5-sonnet
-EOF
-
-# Load and run
-set -a; source .env; set +a
+# OR use direct executable
 codemie-code
-```
 
-#### Method 3: Inline
+# Execute single task
+codemie-code --task "Help me debug this error"
 
-```bash
-CODEMIE_BASE_URL="https://your-ai-proxy.com" \
-CODEMIE_AUTH_TOKEN="your-token" \
-CODEMIE_MODEL="claude-4-5-sonnet" \
-codemie-code
+# With initial message
+codemie-code "Review my code for bugs"
 ```
 
-### Advanced: Provider-Specific Configuration
+#### Install Additional Agents
 
-For advanced scenarios, you can use provider-specific environment variables. These act as fallbacks if generic `CODEMIE_*` variables are not set:
-
-**For Anthropic Claude:**
 ```bash
-export ANTHROPIC_BASE_URL="https://api.anthropic.com"
-export ANTHROPIC_AUTH_TOKEN="your-anthropic-token"
-export ANTHROPIC_MODEL="claude-4-5-sonnet"
-```
+# Install Claude Code
+codemie install claude
 
-**For OpenAI GPT:**
-```bash
-export OPENAI_BASE_URL="https://api.openai.com/v1"
-export OPENAI_API_KEY="your-openai-key"
-export OPENAI_MODEL="gpt-4"
+# Install Codex
+codemie install codex
 ```
 
-**Configuration Priority:**
-1. `CODEMIE_*` (generic - checked first)
-2. `ANTHROPIC_*` or `OPENAI_*` (provider-specific - fallback)
-
----
-
-## 🎯 Usage
-
-### CodeMie Code Assistant
-
-The built-in AI assistant with filesystem, git, and command execution tools.
+#### List Available Agents
 
 ```bash
-# Interactive mode (default)
-codemie-code
-codemie-code /path/to/project
-
-# Execute one-off task
-codemie-code exec "Add error handling to api.ts"
-
-# Execute with custom directory
-codemie-code exec "Run tests" -d /path/to/project
-
-# Test connection
-codemie-code test
-
-# With MCP servers
-codemie-code --mcp-servers time,context7
-
-# With specific model
-CODEMIE_MODEL="claude-opus" codemie-code
-```
-
-#### Interactive Commands
-
-Once inside the assistant:
-
-- **exit** - Exit the assistant
-- **clear** - Clear conversation history
-- Any coding question or task
-
-#### Example Prompts
-
-```
-You: Show me the project structure
-You: Read the package.json file
-You: List all TypeScript files
-You: Show git status and recent commits
-You: Create a new file called utils.ts with helper functions
-You: Search for files containing "authentication"
-You: Edit README.md to add a new section
-You: Run npm test and show me the results
+codemie list
 ```
 
-### MCP Server Management
-
-Extend capabilities with Model Context Protocol servers.
-
-```bash
-# List available servers
-codemie-code mcp list
-
-# List with detailed configuration
-codemie-code mcp list -v
-
-# Add new server
-codemie-code mcp add <name> <command-or-url> [options]
-
-# Test server configuration
-codemie-code mcp test <server-name>
+#### Run an Agent
 
-# Remove server
-codemie-code mcp remove <server-name>
+Agents are run directly after installation. Configuration is automatically passed from `~/.codemie/config.json`.
 
-# Preview which servers will load
-codemie-code mcp servers
-codemie-code mcp servers --servers time,context7
-```
-
-#### Popular MCP Servers
-
-**Time Server** - Query times in any timezone:
 ```bash
-codemie-code mcp add time uvx -a "mcp-server-time"
-codemie-code --mcp-servers time
-
-# Then ask: "What time is it in Hong Kong?"
-```
+# Run built-in CodeMie Native
+codemie run codemie-code
 
-**Context7** - Access up-to-date library documentation:
-```bash
-codemie-code mcp add context7 npx -a "-y" "@upstash/context7-mcp"
-codemie-code --mcp-servers context7
+# Run Claude
+codemie run claude
 
-# Then ask: "Show me LangChain documentation for chains"
+# Run Codex
+codemie run codex
 ```
 
-### CLI Wrapper (codemie)
+**Automatic Model Validation:**
 
-Manage multiple AI coding agents from one interface.
+CodeMie automatically validates model compatibility:
+- **Codex** only accepts OpenAI models (gpt-5, gpt-4.1, gpt-4o, etc.)
+- **Claude** accepts both Claude and GPT models
 
-#### List Available Agents
+If you try to run Codex with a Claude model, CodeMie will:
+1. Detect the incompatibility
+2. **Fetch available models** from your provider's `/v1/models` API endpoint
+3. Filter to show only compatible models
+4. Offer to switch to a compatible GPT model
+5. Optionally save your choice
 
-```bash
-codemie list                    # List all agents
-codemie list --installed        # List only installed agents
-```
+**Dynamic Model Discovery:**
+- Models are fetched in real-time from your configured provider
+- No hardcoded lists - always shows what's actually available
+- Results are cached for 5 minutes to improve performance
+- Works with any OpenAI-compatible API (AI/Run, LiteLLM, OpenAI, etc.)
 
-#### Install Agents
+---
 
-```bash
-codemie install claude-code     # Install Anthropic Claude Code
-codemie install aider           # Install Aider
-codemie install codex           # Install OpenAI Codex
-```
+## 📚 Available Commands
 
-#### Run Agents
+### Core Commands
 
 ```bash
-# Run built-in CodeMie Code
-codemie run codemie-code
-
-# Run other agents
-codemie run claude-code
-codemie run aider
-codemie run codex
+# Setup wizard (interactive configuration)
+codemie setup
 
-# With custom model
-codemie run codemie-code --model claude-opus
+# Environment configuration guide
+codemie env
 
-# Pass additional arguments
-codemie run aider --no-auto-commits
-```
+# List all available agents
+codemie list
 
-#### Uninstall Agents
+# Install an agent
+codemie install <agent>
 
-```bash
-codemie uninstall claude-code
-codemie uninstall aider
-```
+# Uninstall an agent
+codemie uninstall <agent>
 
-#### Health Check
+# Run an agent
+codemie run <agent>
 
-```bash
+# Check installation and configuration
 codemie doctor
-```
 
-Checks:
-- ✅ Node.js version (>= 24.0.0)
-- ✅ npm installation
-- ✅ git availability
-- ✅ Environment variables configuration
-- ✅ Installed agents
+# Manage configuration
+codemie config [options]
 
-#### Version Information
-
-```bash
+# Show version
 codemie version
-codemie-code --version
 ```
 
----
-
-## 🛠️ Available Tools
-
-CodeMie Code assistant comes with 15 built-in tools, plus extensible MCP tools:
-
-### Filesystem Tools (10 tools)
-
-| Tool | Description |
-|------|-------------|
-| `read_file` | Read contents of a file |
-| `read_multiple_files` | Read multiple files at once |
-| `write_file` | Write/create a file with content |
-| `edit_file` | Edit file with diff-based updates |
-| `create_directory` | Create directories recursively |
-| `list_directory` | List files and directories |
-| `project_tree` | Generate visual project structure tree |
-| `move_file` | Move or rename files |
-| `search_files` | Search files by name or content |
-| `list_allowed_directories` | Show accessible directories |
-
-### Git Tools (4 tools)
-
-| Tool | Description |
-|------|-------------|
-| `git_status` | Get repository status |
-| `git_diff` | Show git diff for changes |
-| `git_log` | Show commit history |
-| `git_command` | Execute any git command |
-
-### Command Execution (1 tool)
-
-| Tool | Description |
-|------|-------------|
-| `execute_command` | Execute shell commands (with security checks) |
+### Built-in Agent Commands
 
-### MCP Tools (Extensible)
-
-MCP (Model Context Protocol) enables dynamic tool loading from external servers:
-
-| Server | Tools | Description |
-|--------|-------|-------------|
-| `context7` | `resolve-library-id`, `get-library-docs` | Access up-to-date library documentation |
-| `time` | `get-current-time` | Query times in any timezone |
-| Custom | Varies | Add your own MCP servers |
-
-Configure MCP servers in `~/.codemie/config.json` to extend available tools.
-
-### Tool Features
-
-- **Path Validation**: All operations restricted to allowed directories
-- **Symlink Protection**: Prevents escaping allowed directories
-- **Ignore Patterns**: Automatically excludes node_modules, .git, etc.
-- **Indentation Preservation**: Smart editing maintains code formatting
-- **Unified Diffs**: Shows clear before/after changes
-- **Command Sanitization**: Blocks dangerous commands (rm -rf /, sudo, etc.)
-
----
-
-## 🔒 Security Features
+```bash
+# CodeMie Native - Interactive mode
+codemie-code
 
-CodeMie Code implements multiple security layers:
+# CodeMie Native - Single task execution
+codemie-code --task "Your task here"
 
-### Filesystem Security
+# CodeMie Native - With initial message
+codemie-code "Your initial message"
 
-✅ **Path Validation**
-- All paths validated against allowed directories
-- No directory traversal attacks (../)
-- Absolute path resolution
+# CodeMie Native - Health check
+codemie-code health
 
-✅ **Symlink Protection**
-- Symlinks resolved and validated
-- Target paths must be within allowed directories
+# CodeMie Native - Debug mode
+codemie-code --debug
+```
 
-✅ **Ignore Patterns**
-- Automatically excludes sensitive directories
-- Default: `node_modules`, `.git`, `__pycache__`, `.venv`, `dist`, `build`
+### Direct Agent Shortcuts
 
-### Command Security
+Quick access to external agents with configuration overrides:
 
-The `execute_command` tool blocks dangerous patterns:
+```bash
+# Claude Code (direct shortcut)
+codemie-claude                              # Interactive mode
+codemie-claude --model claude-4-5-sonnet    # Override model
+codemie-claude --api-key your-key           # Override API key
+codemie-claude health                       # Health check
 
-```typescript
-🚫 rm -rf /              // Recursive delete on root
-🚫 mkfs                  // Filesystem formatting
-🚫 dd if=                // Disk operations
-🚫 wget ... | sh         // Download and execute
-🚫 curl ... | sh         // Download and execute
-🚫 sudo                  // Privilege escalation
-🚫 chmod 777             // Unsafe permissions
-🚫 > /etc/               // Writing to system config
-🚫 > /dev/               // Writing to devices
-🚫 :(){ :|:& };:         // Fork bomb
+# Codex (direct shortcut)
+codemie-codex                               # Interactive mode
+codemie-codex --model gpt-4o                # Override model (OpenAI only)
+codemie-codex --provider openai             # Override provider
+codemie-codex health                        # Health check
 ```
 
-### Security Best Practices
-
-1. **Set Allowed Directories**: Only the working directory is accessible
-2. **Review Changes**: Always review diffs before applying
-3. **Use Version Control**: Work in git repositories
-4. **Test in Isolation**: Test in separate environments first
-
----
-
-## 🎨 Usage Examples
+**Features of Direct Shortcuts:**
+- Bypass the registry system for faster startup
+- Support all original agent options and arguments
+- Allow configuration overrides via CLI flags
+- Include health check commands
+- Pass through unknown options to the underlying agent
 
-### Example 1: Code Review
+### Configuration Commands
 
 ```bash
-codemie-code
+# Set a configuration value
+codemie config set <key> <value>
 
-You: Read src/utils/helper.ts and review the code
-Assistant: [Reads file, provides detailed review with suggestions]
+# Get a configuration value
+codemie config get <key>
 
-You: Create a test file for this module
-Assistant: [Creates tests/utils/helper.test.ts with test cases]
-```
-
-### Example 2: Refactoring
-
-```bash
-You: Show me all files in src/components
-Assistant: [Lists component files]
+# List all configuration
+codemie config list
 
-You: Read Button.tsx and refactor it to use hooks
-Assistant: [Shows diff with proposed changes]
+# Reset configuration
+codemie config reset
 
-You: Apply those changes
-Assistant: [Applies edits and shows confirmation]
+# Advanced: Change timeout (default: 300 seconds)
+codemie config set timeout 600
 ```
 
-### Example 3: Git Workflow
+---
 
-```bash
-You: Show me the current git status
-Assistant: [Shows modified, staged, and untracked files]
+## ⚙️ Configuration
 
-You: Show diff for src/app.ts
-Assistant: [Displays git diff]
+CodeMie stores configuration in `~/.codemie/config.json`:
 
-You: Show the last 5 commits
-Assistant: [Shows commit history]
+```json
+{
+  "provider": "anthropic",
+  "model": "claude-4-5-sonnet",
+  "anthropic": {
+    "baseUrl": "https://api.anthropic.com/v1",
+    "apiKey": "your-api-key",
+    "timeout": 300
+  }
+}
 ```
 
-### Example 4: Project Setup
+### Supported Providers
 
-```bash
-You: Create a new directory called "api"
-Assistant: [Creates directory]
+- **anthropic** - Anthropic Claude API
+- **openai** - OpenAI API
+- **azure** - Azure OpenAI
+- **bedrock** - AWS Bedrock
+- **litellm** - LiteLLM Proxy (custom)
 
-You: Create index.ts in api/ with a basic Express setup
-Assistant: [Creates file with boilerplate code]
+### Environment Variables
 
-You: Create a README.md in api/ documenting the API
-Assistant: [Creates documentation]
-```
-
-### Example 5: Search and Replace
+You can also configure via environment variables (higher priority than config file):
 
 ```bash
-You: Search for all files containing "TODO"
-Assistant: [Lists files with TODO comments]
+# Anthropic
+export ANTHROPIC_BASE_URL="https://api.anthropic.com/v1"
+export ANTHROPIC_API_KEY="your-api-key"
+export ANTHROPIC_MODEL="claude-4-5-sonnet"
 
-You: Read components/Header.tsx
-Assistant: [Shows file content]
+# OpenAI
+export OPENAI_BASE_URL="https://api.openai.com/v1"
+export OPENAI_API_KEY="your-api-key"
+export OPENAI_MODEL="gpt-4"
 
-You: Replace the TODO with an implementation
-Assistant: [Shows diff and applies changes]
+# Generic (works with any provider)
+export AI_BASE_URL="https://your-proxy.com"
+export AI_API_KEY="your-api-key"
+export AI_MODEL="your-model"
 ```
 
 ---
 
-## 🐛 Troubleshooting
-
-### Common Issues
+## 🤖 Supported Agents
 
-#### "Base URL not configured"
+### CodeMie Native (Built-in) ⭐
 
-**Solution:**
-```bash
-export CODEMIE_BASE_URL="https://your-ai-proxy.com"
-export CODEMIE_AUTH_TOKEN="your-token"
-export CODEMIE_MODEL="claude-4-5-sonnet"
-```
-
-#### "Connection test failed"
-
-**Possible causes:**
-1. Invalid CODEMIE_AUTH_TOKEN
-2. CODEMIE_BASE_URL not accessible
-3. Network/VPN issues
+**NEW:** CodeMie's built-in LangGraph-based coding assistant. No external installation required!
 
-**Solution:**
+**Usage:**
 ```bash
-# Verify URL is accessible
-curl -I $CODEMIE_BASE_URL
+# Interactive mode
+codemie run codemie-code
 
-# Check environment variables
-echo $CODEMIE_BASE_URL
-echo $CODEMIE_AUTH_TOKEN
-echo $CODEMIE_MODEL
+# Direct executable (alternative)
+codemie-code
 
-# Run doctor
-codemie doctor
+# Single task execution
+codemie-code --task "Fix the bug in utils.js"
 
-# Try test again
-codemie-code test
+# With initial message
+codemie-code "Help me refactor this component"
 ```
 
-#### "Command not found: codemie"
+**Features:**
+- 🔧 **Built-in**: No installation required - ready to use immediately
+- 🚀 **Modern UI**: Beautiful terminal interface with real-time streaming
+- 🎯 **Task-focused**: Execute single tasks or start interactive sessions
+- 🔍 **Debug Mode**: Comprehensive logging with `--debug` flag
+- 📋 **Clipboard Support**: Automatic image detection from clipboard
+- 🛠️ **System Tools**: File operations, git integration, and command execution
+- 💬 **Interactive Chat**: Continuous conversations with context memory
+- 📊 **Usage Statistics**: Token tracking and cost monitoring
 
-**Solution:**
+**Direct Commands:**
 ```bash
-# Re-install globally
-npm install -g @codemieai/code
+# Health check
+codemie-code health
 
-# Or re-link the package (from source)
-cd /path/to/codemie-code
-npm link
+# Interactive mode with debug
+codemie-code --debug
 
-# Verify
-which codemie
-which codemie-code
+# Execute task and exit
+codemie-code --task "Run tests and fix any failures"
 ```
 
-#### "Path outside allowed directories"
+### Claude Code
 
-**Solution:**
-CodeMie Code only operates within the working directory for security.
+Anthropic's official CLI for Claude AI. Provides advanced code understanding, generation, and refactoring capabilities.
 
+**Installation:**
 ```bash
-# Start in the correct directory
-cd /path/to/your/project
-codemie-code
-
-# Or specify the directory
-codemie-code /path/to/your/project
+codemie install claude
 ```
 
-#### Environment Variables Not Loaded
+**Features:**
+- Advanced code understanding
+- Multi-file editing
+- Interactive conversations
+- Project-aware context
 
-```bash
-# Verify variables are set
-env | grep CODEMIE_
+### Codex
 
-# If empty, reload shell profile
-source ~/.bashrc  # or ~/.zshrc
-
-# Or export again
-export CODEMIE_BASE_URL="..."
-export CODEMIE_AUTH_TOKEN="..."
-export CODEMIE_MODEL="..."
-```
-
-### Debug Mode
-
-Enable debug logging to diagnose issues:
+OpenAI's code generation assistant. Optimized for code completion and generation tasks.
 
+**Installation:**
 ```bash
-export CODEMIE_DEBUG="true"
-codemie-code
+codemie install codex
 ```
 
-Debug mode shows:
-- Tool initialization details
-- AI request/response info
-- File operation details
-- Error stack traces
-
-### Clean Installation
-
-If you encounter persistent issues:
+**Features:**
+- Code completion
+- Function generation
+- Bug fixing
+- Code explanation
 
-```bash
-# Clean build
-rm -rf dist/ node_modules/
-npm install
-npm run build
+**Model Compatibility:**
+- ✅ Supports: GPT models (gpt-5, gpt-5-codex, gpt-4.1, gpt-4o)
+- ❌ Not supported: Claude/Anthropic models (API incompatibility)
 
-# Re-link
-npm link
-```
+> **Note:** Codex requires OpenAI-compatible models. If you configure a Claude model, CodeMie will automatically prompt you to switch to a compatible GPT model.
 
 ---
 
-## 📊 Architecture
+## 🛠️ Development
 
-### Package Structure
+### Project Structure
 
 ```
 codemie-code/
 ├── bin/
-│   ├── codemie-code.js      # AI assistant entry point
-│   └── codemie.js           # CLI wrapper entry point
-│
+│   └── codemie.js               # CLI entry point
 ├── src/
-│   ├── code/                # CodeMie Code Assistant
-│   │   ├── index.ts         # Main assistant class
-│   │   ├── agent.ts         # LangChain ReAct agent
-│   │   ├── agent-events.ts  # Event system for streaming
-│   │   ├── config.ts        # Configuration loader
-│   │   ├── prompts.ts       # System prompts
-│   │   └── tools/           # Tool implementations
-│   │       ├── filesystem.ts   # 10 filesystem tools
-│   │       ├── git.ts          # 4 git tools
-│   │       ├── command.ts      # Command execution
-│   │       ├── mcp.ts          # MCP integration
-│   │       └── diff-utils.ts   # Diff utilities
-│   │
-│   ├── cli/                 # CLI Wrapper
-│   │   └── commands/        # CLI commands
-│   │       ├── list.ts
-│   │       ├── install.ts
-│   │       ├── run.ts
-│   │       ├── doctor.ts
-│   │       ├── uninstall.ts
-│   │       ├── version.ts
-│   │       └── mcp.ts
-│   │
-│   ├── agents/              # Agent System
-│   │   ├── registry.ts      # Agent registry
-│   │   └── adapters/        # Agent adapters
-│   │       ├── codemie-code.ts
-│   │       ├── claude-code.ts
-│   │       ├── codex.ts
-│   │       └── aider.ts
-│   │
-│   ├── env/                 # Environment Management
-│   │   └── manager.ts
-│   │
-│   ├── ui/                  # UI Layer
-│   │   └── terminal-ui.ts   # Interactive terminal
-│   │
-│   └── utils/               # Utilities
-│       ├── env-mapper.ts    # Environment variable mapping
-│       ├── exec.ts          # Process execution
-│       ├── logger.ts        # Logging
-│       ├── errors.ts        # Error classes
-│       └── tips.ts          # Loading tips
-│
-├── tests/                   # Test files
-│   └── integration/         # Integration tests
-│
-├── dist/                    # Build output (TypeScript compilation)
-├── mcp/                     # MCP server configurations
-│   └── servers.json
-├── package.json             # Dependencies and scripts
-├── tsconfig.json            # TypeScript configuration
-├── eslint.config.mjs        # ESLint configuration
-└── README.md                # Package documentation
+│   ├── agents/                  # Agent adapters
+│   │   ├── registry.ts          # Agent registry
+│   │   └── adapters/            # Agent implementations
+│   ├── cli/                     # CLI commands
+│   │   ├── index.ts             # CLI setup
+│   │   └── commands/            # Command implementations
+│   ├── env/                     # Environment management
+│   └── utils/                   # Utilities
+├── package.json
+└── README.md
 ```
 
-### Technology Stack
+### Building
 
-- **Language**: TypeScript (ES2022, NodeNext modules)
-- **Runtime**: Node.js >= 24.0.0 (LTS Krypton)
-- **Package Manager**: npm
-- **LLM Framework**: LangChain 1.x (`@langchain/core`, `@langchain/langgraph`, `@langchain/openai`)
-- **LLM Provider**: Any OpenAI-compatible API
-- **CLI Framework**: Commander.js
-- **Schema Validation**: Zod
-- **Diff Generation**: diff package
-- **UI**: Chalk, Inquirer, Ora
-- **Testing**: Jest with ts-jest
-- **Linting**: ESLint with TypeScript support
+```bash
+npm run build       # Compile TypeScript
+npm run dev         # Watch mode for development
+npm run lint        # Check code style
+npm run lint:fix    # Fix linting issues
+```
 
----
+### Testing
 
-## 🔄 Roadmap
-
-### Current Status (v1.0.0)
-
-✅ **Implemented**:
-- Built-in AI coding assistant
-- CLI wrapper for multiple agents
-- 10 filesystem tools with security
-- 4 git operation tools
-- 1 command execution tool
-- MCP (Model Context Protocol) integration
-- Diff-based file editing
-- Agent management (install/uninstall/run)
-- Health check and diagnostics
-- Interactive terminal UI with cancellation support
-- One-off task execution with `exec` command
-
-### Planned Features (v1.1.0+)
-
-- 🎨 Custom themes and output formats
-- 📦 Enhanced plugin system for custom tools
-- 🌐 Multi-language support
-- 📊 Usage analytics and insights
-- 🔌 VS Code extension integration
-- 🤖 More agent adapters (Cursor, Copilot, etc.)
-- 🔍 Advanced MCP server discovery and management
+After building, test the CLI locally:
 
----
+```bash
+# Build and link
+npm run build && npm link
 
-## 🤝 Contributing
+# Test commands
+codemie --help
+codemie list
+codemie doctor
+```
 
-Contributions are welcome! Please:
+---
 
-1. Fork the repository
-2. Create a feature branch
-3. Make your changes
-4. Add tests if applicable
-5. Submit a pull request
+## 🐛 Troubleshooting
 
-### Development Setup
+### Command not found after installation
 
+Re-link the package:
 ```bash
-# Clone repository
-git clone https://github.com/codemie-ai/codemie-code.git
-cd codemie-code
-
-# Install dependencies
-npm install
-
-# Build
-npm run build
-
-# Watch mode (for development)
-npm run dev
-
-# Test locally
 npm link
-codemie doctor
+which codemie
 ```
 
-### Running Tests
+### Configuration not found
 
+Run the setup wizard:
 ```bash
-# Run all tests
-npm test
+codemie setup
+```
 
-# Run with coverage
-npm run test:coverage
+Or check your configuration:
+```bash
+codemie config list
+```
 
-# Run specific test
-npm test -- path/to/test
+### Agent installation fails
 
-# Manual integration tests
-node tests/test-streaming.js
-node tests/test-agent-direct.js
+Check your internet connection and try again:
+```bash
+codemie install <agent>
 ```
 
----
-
-## 📄 License
+### Connection issues
 
-MIT License - see LICENSE file for details
+Verify your configuration:
+```bash
+codemie doctor
+```
 
 ---
 
-## 🙏 Acknowledgments
+## 📄 License
 
-- Built with [LangChain](https://js.langchain.com/)
-- Inspired by [Aider](https://github.com/paul-gauthier/aider) and [Claude Code](https://claude.com/code)
-- CLI framework by [Commander.js](https://github.com/tj/commander.js)
-- MCP specification by [Anthropic](https://github.com/modelcontextprotocol/specification)
+Apache-2.0
 
 ---
 
-## 📞 Support
+## 🤝 Contributing
 
-For issues, questions, or contributions:
-- 🐛 Report bugs via [GitHub Issues](https://github.com/codemie-ai/codemie-code/issues)
-- 💬 Ask questions in [Discussions](https://github.com/codemie-ai/codemie-code/discussions)
-- 📖 Read the full [User Guide](docs/USER_GUIDE.md)
+Contributions are welcome! Please read our contributing guidelines before submitting PRs.
 
 ---
 
-## Quick Reference
-
-### Essential Commands
-
-```bash
-# Setup
-export CODEMIE_BASE_URL="https://your-ai-proxy.com"
-export CODEMIE_AUTH_TOKEN="your-token"
-export CODEMIE_MODEL="claude-4-5-sonnet"
-
-# Verify
-codemie doctor
-codemie-code test
-
-# Start coding
-codemie-code
-codemie-code /path/to/project
-codemie-code exec "task description"
-
-# MCP
-codemie-code mcp list
-codemie-code --mcp-servers time,context7
-
-# CLI Wrapper
-codemie list
-codemie install claude-code
-codemie run claude-code
-```
-
-### Common Tasks
-
-| What you want | Command |
-|---------------|---------|
-| Start interactive coding | `codemie-code` |
-| Execute single task | `codemie-code exec "task"` |
-| Test connection | `codemie-code test` |
-| Check system health | `codemie doctor` |
-| List MCP servers | `codemie-code mcp list` |
-| Install agent | `codemie install <agent>` |
-| Show versions | `codemie version` |
-
----
+## 🔗 Links
 
-**Made with ❤️ by the CodeMie Team**
+- [GitHub Repository](https://github.com/codemie-ai/codemie-code)
+- [Issue Tracker](https://github.com/codemie-ai/codemie-code/issues)
+- [NPM Package](https://www.npmjs.com/package/@codemieai/code)