OpenAgent CLI

Intelligent agentic coder with GraphRAG engine, MCP integration, and multi-LLM support

Features

🤖 Multi-LLM Support: OpenAI, Anthropic, local LLM Studio integration
📊 GraphRAG Engine: Intelligent codebase understanding and memory
🔧 MCP Integration: Model Context Protocol for tool execution
⚡ Streaming Responses: Real-time AI responses
🎯 Multi-format Tool Parsing: Custom, XML, JSON tool call formats
🧠 Dynamic System Prompts: Auto-discovery of available tools
📁 Filesystem Tools: Read, write, search files and directories

Installation

npm install -g openagent-cli

Quick Start

1. Initialize configuration:

openagent config --init

2. Start the interactive UI:

openagent
# or
openagent ui

3. Configure your LLM provider in config.json:

{
  "providers": [
    {
      "name": "local",
      "type": "openai", 
      "baseURL": "http://localhost:1234/v1",
      "apiKey": "lm-studio",
      "defaultModel": "your-model-name"
    }
  ]
}

Configuration

Example config.json

{
  "providers": [
    {
      "name": "openai",
      "type": "openai",
      "apiKey": "your-openai-key",
      "defaultModel": "gpt-4"
    },
    {
      "name": "local",
      "type": "openai",
      "baseURL": "http://localhost:1234/v1", 
      "apiKey": "lm-studio",
      "defaultModel": "deepseek/deepseek-r1-0528-qwen3-8b"
    }
  ],
  "mcpServers": [
    {
      "name": "fs",
      "command": "npx",
      "args": ["@modelcontextprotocol/server-filesystem", "."]
    }
  ],
  "agents": [
    {
      "id": "assistant",
      "provider": "local",
      "mcpTools": ["fs"],
      "system": "You are an intelligent AI assistant..."
    }
  ]
}

Commands

Interactive UI (Default)

openagent            # Start interactive UI
openagent ui         # Same as above
openagent ui -c /path/to/config.json

Codebase Indexing (Coming Soon)

openagent index /path/to/codebase
openagent index . --languages typescript,python --parallel 4

Query Codebase (Coming Soon)

openagent query "how does authentication work?"
openagent query "find all API endpoints" --limit 20

HTTP Server (Coming Soon)

openagent server --port 3001 --websocket

Configuration Management

openagent config --init        # Create default config
openagent config --validate    # Validate config file  
openagent config --show        # Show current config

Tool Usage

OpenAgent supports multiple tool call formats:

Primary Format (Recommended)

[TOOL_REQUEST]
{"name": "write_file", "arguments": {"path": "hello.txt", "contents": "Hello World"}}
[END_TOOL_REQUEST]

Alternative Formats

<tool_call name="read_file" args='{"path": "hello.txt"}'></tool_call>

{"tool": "list_directory", "args": {"path": "."}}

Available Tools

Filesystem Tools

• write_file - Create or overwrite files
• read_file - Read file contents
• list_directory - List files and folders
• create_directory - Create directories
• move_file - Move/rename files
• search_files - Search for files matching patterns

Built-in Tools

• echo - Echo a message
• timestamp - Get current timestamp
• random - Generate random numbers
• math - Perform calculations

Development

Local Development

# Clone repository  
git clone https://github.com/yourusername/openagent.git
cd openagent

# Install dependencies
npm install

# Start development UI
npm run dev

# Build
npm run build

# Run tests
npm test

Project Structure

src/
├── cli.ts              # Command line interface
├── ui.tsx              # Interactive UI component  
├── main.ts             # Core application logic
├── config.ts           # Configuration management
├── providers.ts        # LLM provider integrations
├── mcp.ts              # MCP server management
├── tools/              # Tool system
│   ├── tool-parser.ts  # Multi-format tool parsing
│   ├── tool-executor.ts # Tool execution engine
│   └── system-prompt-builder.ts # Dynamic prompt generation
└── ...

Roadmap

Phase 1: Core Infrastructure ✅

• Multi-LLM provider support
• MCP server integration
• Streaming responses
• Multi-format tool parsing
• Dynamic system prompts

Phase 2: GraphRAG Engine (In Progress)

• PostgreSQL vector database
• Codebase indexing system
• Semantic search capabilities
• Knowledge graph construction

Phase 3: Advanced Features (Planned)

• HTTP streaming MCP server
• Code pattern learning
• Automated refactoring suggestions
• Multi-repository support

Contributing

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Add tests if applicable
5. Submit a pull request

License

MIT License - see LICENSE file for details.

Support

• 🐛 Report Issues
• 💬 Discussions
• 📖 Documentation

---

Made with ❤️ by the OpenAgent community