Model Context Protocol (MCP) is like a USB for AI tools - it provides a standardized way for AI agents to discover and use different capabilities. Think of it as plugging in new abilities for your AI agents! With our flexible dictionary-based implementation, agents can easily discover and use tools without needing to know their implementation details.

The MCP ecosystem is growing rapidly with many powerful servers available:

• Brave Search - Web and local search capabilities
• Filesystem - Complete file and directory operations
• Playwright - Browser automation and web scraping
• MongoDB - Direct database interactions
• Sequential Thinking - Structured reasoning and planning
• FLUX - Image generation and manipulation
• Selenium - Web automation and testing
• SQLite - Local database operations
• Vector Store - Embedding and similarity search
• And many more!

Check out the MCP Servers Repository for the latest list of available servers. The ecosystem is rapidly growing with new capabilities being added regularly!

1. Install Requirements:

# Create and activate virtual environment
python -m venv .env
source .env/bin/activate  # On Windows: .env\Scripts\activate

# Install required packages
pip install mcp

2. Configure MCP:

# Copy the example config
cp mcp_config.example.json mcp_config.json

# Edit mcp_config.json with your settings
{
  "mcpServers": {
    "brave-search": {
      "enabled": true,
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-brave-search"],
      "env": {
        "BRAVE_API_KEY": "your-api-key-here"
      }
    },
    "filesystem": {
      "enabled": true,
      "command": "npx",
      "args": [
        "-y",
        "@modelcontextprotocol/server-filesystem",
        "/path/to/your/working/directory"
      ]
    }
  }
}

3. Start AutoGen Studio:

autogenstudio ui --port 8080

Our implementation uses a flexible dictionary-based approach, making it easy to discover and use tools:

# 1. List Available Servers
servers = await mcp(tool='list_available_servers')

# 2. Discover Server Tools
tools = await mcp(
    server='brave-search',
    tool='tool_details'
)

# 3. Use Tools with Dictionary Arguments
# Web Search Example
result = await mcp(
    server='brave-search',
    tool='brave_web_search',
    arguments={
        'query': 'Latest AI developments',
        'count': 5
    }
)

# File Operations Example
result = await mcp(
    server='filesystem',
    tool='read_file',
    arguments={
        'path': '/path/to/your/file'
    }
)

We provide ready-to-use templates to help you get started:

1. Gallery Examples (/gallerys):

   • Pre-configured tool setups
   • Example implementations
   • Best practices demonstrations

2. Team Templates (/teams):

   • Complete agent workflows
   • Tool integration examples
   • Task-specific configurations

To use a template:

1. Copy the desired template from /gallerys or /teams
2. Customize the JSON configuration for your needs
3. Load it in AutoGen Studio

• Web search capabilities
• Local business search
• Configurable result limits
• Fresh content filtering

• Read/Write files
• Directory operations
• File searches
• Metadata operations

The MCP ecosystem is constantly growing. Add new servers to your config to expand capabilities!

The MCP client checks these locations for config files:

1. Current directory: ./mcp_config.json
2. User config: ~/.config/autogen/mcp_config.json
3. Environment: $MCP_CONFIG_PATH

Run the test suite:

pytest tests/test_mcp_client.py -v
pytest tests/test_mcp_json.py -v

The /mcp_workspace directory is provided for testing but ignored by git. Your tests will create this automatically.

• Python 3.12+
• Node.js and npx
• MCP Python SDK
• AutoGen Studio

MIT License - See LICENSE file for details