◈ OmniAgent — Universal AI CLI Agent

A powerful, beautiful CLI agent that works with any AI provider and has full filesystem + shell capabilities.

  ██████╗ ███╗   ███╗███╗   ██╗██╗     ███████╗ ██████╗ ███████╗███╗   ██╗████████╗
 ██╔═══██╗████╗ ████║████╗  ██║██║     ██╔════╝██╔════╝ ██╔════╝████╗  ██║╚══██╔══╝
 ██║   ██║██╔████╔██║██╔██╗ ██║██║     ███████╗██║  ███╗█████╗  ██╔██╗ ██║   ██║   
 ██║   ██║██║╚██╔╝██║██║╚██╗██║██║     ╚════██║██║   ██║██╔══╝  ██║╚██╗██║   ██║   
 ╚██████╔╝██║ ╚═╝ ██║██║ ╚████║██║     ███████║╚██████╔╝███████╗██║ ╚████║   ██║   
  ╚═════╝ ╚═╝     ╚═╝╚═╝  ╚═══╝╚═╝     ╚══════╝ ╚═════╝ ╚══════╝╚═╝  ╚═══╝   ╚═╝   

🚀 Quick Start

# Install dependencies
npm install

# First run — guided setup wizard
node index.js

# Or set up separately
node index.js --setup

🔌 Supported Providers

| Provider | Format | Notes | |----------|--------|-------| | Anthropic (Claude) | Native | claude-opus, sonnet, haiku | | OpenAI | OpenAI | gpt-4o, o1, o3-mini | | Groq | OpenAI | Ultra fast inference | | xAI (Grok) | OpenAI | grok-2, grok-beta | | DeepSeek | OpenAI | deepseek-chat, reasoner | | Mistral AI | OpenAI | mistral-large | | Together AI | OpenAI | Llama, Qwen, etc. | | OpenRouter | OpenAI | All models via one key | | io.net | OpenAI | Decentralized inference | | Ollama | OpenAI | Local models | | Custom | OpenAI | Any OpenAI-compatible API (kiai.ai, etc.) |

💻 Usage

# Start in current directory
node index.js

# Start in specific directory
node index.js /path/to/project

# Global install (optional)
npm link
omni-agent /path/to/project

⌨️ Commands

| Command | Description | |---------|-------------| | /help | Show all commands | | /clear | Clear screen + conversation history | | /stats | Show current session statistics | | /cd <dir> | Change working directory | | /model <name> | Switch model on the fly | | /setup | Reconfigure provider/API key/model | | /reset | Start new conversation (clear history) | | /history | Show conversation history | | /exit | Exit and show session statistics |

🛠️ Agent Tools

The agent has access to 13 filesystem and shell tools:

| Tool | Description | |------|-------------| | list_directory | List files with metadata | | read_file | Read file content | | write_file | Create/overwrite files | | append_to_file | Append to files | | patch_file | Surgically edit text in files | | delete_path | Delete files/directories | | copy_path | Copy files | | move_path | Move/rename files | | create_directory | Create directories recursively | | get_file_info | Get file metadata | | find_files | Find files by glob pattern | | search_in_files | Grep search in files | | execute_command | Run any shell command |

⚡ Token Efficiency

OmniAgent uses parallel tool calling — when the AI needs multiple independent pieces of information, it fetches them ALL in a single API call instead of making sequential requests.

Example: "List the src/ folder and read package.json"

• ❌ Naive: 3 API calls (list → read → answer)
• ✅ OmniAgent: 2 API calls (list+read simultaneously → answer)

📊 Session Statistics

On exit (or /stats), OmniAgent shows:

• Session duration
• Messages sent/received
• API requests made
• Input/output token counts
• Estimated cost
• Tool usage breakdown with visual bars

⚙️ Config File

Config is saved to ~/.omni-agent.json:

{
  "providerKey": "groq",
  "baseURL": "https://api.groq.com/openai/v1",
  "apiKey": "your-key-here",
  "model": "llama-3.3-70b-versatile",
  "format": "openai",
  "maxTokens": 8192,
  "temperature": 0.3
}

💡 Example Prompts

⟩ Refactor all the .js files in src/ to use async/await instead of callbacks

⟩ Find all TODO comments in this project and create a TODO.md file

⟩ Run the tests and fix any failures

⟩ Create a REST API with Express for a simple todo app

⟩ Analyze this codebase and suggest improvements

⟩ Set up a Python virtual environment and install requirements.txt

Built with Node.js · chalk · gradient-string · figlet · ora · inquirer · boxen