nixagent 1.1

A sophisticated AI agent toolkit supporting multiple AI providers with tool calling capabilities.

Local Agent Toolkit

A generic, multipurpose Local Agent Toolkit in Python. This framework is completely agnostic to specific use cases and architectures, serving as a robust foundation for building autonomous, collaborative AI agents that can manage their own context, interface with each other, and securely use external tools.

🚀 Quick Start

Installation

pip install -r requirements.txt

Command Line Usage

First, set up your environment configuration by copying .env.example to .env and adding your API keys.

# Ask a question directly
python app.py "What files are in the current directory?"

# Interactive mode
python app.py

# With custom settings
python app.py "Analyze the code structure" --no-save

Python Library Usage

from local_agent_toolkit import Agent

# Initialize the core agent
agent = Agent(
    name="MainAgent",
    system_prompt="You are a highly capable AI assistant that uses available tools to accomplish goals."
)

result = agent.run(user_prompt="List all Python files in the project")
print(result)

✨ Features

• 🌐 Standardized API Interface: Uses pure requests following the OpenAI native JSON structure. Compatible with OpenAI, Vertex, Local LLMs (via Ollama/vLLM), Groq, and more.
• 🤖 Autonomous Agents: Agents maintain independent conversation histories and automatically delegate sub-tasks when needed.
• 🔌 Model Context Protocol (MCP): Dynamic tool extension via MCP Servers via .mcp.json.
• 🛠️ Rich Built-In Tools: Deep system-level tools covering regex-based file searching, exact content mapping, disk manipulation, and secure subprocess execution.
• 🗣️ Inter-Agent Collaboration: Support for multiple sub-agents operating concurrently under the same framework via .register_collaborator(agent).

📦 Project Structure

framework/
├── app.py                # Main CLI application
├── local_agent_toolkit/  # Core Framework Mechanics
│   ├── __init__.py       # Library exports
│   ├── agent.py          # Core contextual autonomous Agent
│   ├── llm.py            # Central HTTP-based LLM orchestration
│   ├── mcp.py            # Model Context Protocol definition and bindings
│   └── tools/            # Default Native Tools
│       ├── __init__.py   # Tool bindings & descriptions
│       ├── cmd.py        # Subprocess shell extensions
│       └── fs.py         # File system native operations
├── mcp.json              # Model Context Protocol Server mapping
├── docs/                 # Additional Documentation
├── requirements.txt      # Python dependencies
├── .env                  # Operational mapping variables
└── README.md             # This file

⚙️ Configuration

Create a .env file in your project root:

# Core API Key for the provider (OpenAI, Gemini, Vertex, Groq, etc)
API_KEY=your_api_key_here

# Base URL for the OpenAI compatible endpoint
API_BASE_URL=https://api.openai.com/v1

# The model name to use for generating text
MODEL=gpt-4o

# Tool and Processing Configuration
MAX_ITERATIONS=25

# Logging Configuration
LOG_LEVEL=INFO

🔌 Using MCP Servers

Add server definitions to your mcp.json file in the root directory:

{
  "mcpServers": {
    "sqlite": {
      "command": "uvx",
      "args": ["mcp-server-sqlite", "--db-path", "./database.db"],
      "active": true
    }
  }
}

The framework's MCPManager automatically bootstraps all active MCP servers, parses their schemas, and loads their tools natively alongside standard tools upon Agent initialization.

🤝 Collaborative Agents

Agents can securely establish communication networks.

from local_agent_toolkit import Agent

research_agent = Agent("Researcher", "You perform file system research.")
writer_agent = Agent("Writer", "You answer questions accurately.")

writer_agent.register_collaborator(research_agent)

writer_agent.run("Ask the Researcher to find all text files and read them to me.")

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.