OmniTaskAgent

A multi-model agent for managing tasks across various platforms, requiring API keys for different AI models.

GitHub

OmniTaskAgent

A powerful multi-model task management system that can connect to various task management systems and help users choose and use the task management solution that best suits their needs.

Features

  • Task Management System: Create, list, update and delete tasks, support status tracking and dependency management
  • Task Decomposition and Analysis: Break down complex tasks into subtasks, support complexity assessment and PRD automatic parsing
  • Python Native Implementation: Built entirely in Python, seamlessly integrated with the Python ecosystem
  • Multi-Model Support: Compatible with multiple models like OpenAI, Claude, etc., not limited to specific API providers
  • Editor Integration: Integrate with editors like Cursor through MCP protocol for smooth development experience
  • Intelligent Workflow: Implement intelligent task management process based on LangGraph's ReAct pattern
  • Multi-System Integration: Can connect to various professional task management systems like mcp-shrimp-task-manager and claude-task-master
  • Cross-Scenario Application: Suitable for general development projects, vertical domain projects, and other task systems

Installation

# Install using uv (recommended)
uv pip install -e .

# Or install using pip
pip install -e .

# Install Node.js dependencies (for MCP server)
npm install

Configuration

Create a .env file in the project root directory for configuration:

# Required: API keys (configure at least one)
OPENAI_API_KEY=your_openai_api_key_here
# Or
ANTHROPIC_API_KEY=your_anthropic_api_key_here

# Optional: Model configuration
LLM_MODEL=gpt-4o  # Default model
TEMPERATURE=0.2   # Creativity parameter
MAX_TOKENS=4000   # Maximum tokens

Usage

Command Line Interface (Recommended)

The simplest way to use is through the built-in command line interface:

# Start interactive command line interface
python -m omni_task_agent.cli

Common command examples:

  • Create task: Optimize website performance Reduce page load time by 50%
  • List all tasks
  • Update task 1 status to completed
  • Decompose task 2
  • Analyze project complexity

Using in LangGraph Studio

LangGraph Studio is a development environment specifically designed for LLM applications, used for visualizing, interacting with, and debugging complex agent applications.

First, ensure langgraph-cli is installed (requires version 0.1.55 or higher):

# Install langgraph-cli (requires Python 3.11+)
pip install -U "langgraph-cli[inmem]"

Then start the development server in the project root directory (containing langgraph.json):

# Start local development server
langgraph dev

This will automatically open a browser and connect to the cloud-hosted Studio interface, where you can:

  1. Visualize your agent graph structure
  2. Test and run agents through the UI interface
  3. Modify agent state and debug
  4. Add breakpoints for step-by-step agent execution
  5. Implement human-machine collaboration processes

When modifying code during development, Studio will update automatically without needing to restart the service, facilitating rapid iteration and debugging.

For advanced features like breakpoint debugging:

# Enable debug port
langgraph dev --debug-port 5678

Editor Integration (MCP Service)

  1. Run the MCP server:
# Start STDIO-based MCP service
python run_mcp.py
  1. Configure MCP settings in your editor (like Cursor, VSCode, etc.):
{
  "mcpServers": {
    "task-master-agent": {
      "type": "stdio",
      "command": "/path/to/python",
      "args": ["/path/to/run_mcp.py"],
      "env": {
        "OPENAI_API_KEY": "your-key-here"
      }
    }
  }
}

Project Structure

omnitaskagent/
├── omni_task_agent/     # Main code package
│   ├── agent.py           # LangGraph agent definition
│   ├── config.py          # Configuration management
│   └── cli.py             # Command line interface
├── examples/              # Example code
│   └── basic_usage.py     # Basic usage example
├── tests/                 # Test cases
├── run_mcp.py             # MCP service entry
├── adapters.py            # MCP adapters
├── langgraph.json         # LangGraph API configuration
├── package.json           # Node.js dependencies
└── pyproject.toml         # Python dependencies

Reference Projects

License

MIT

関連サーバー