MCP Agent Team

A multi-agent system for collaborative development using the MCP (Model Context Protocol).

Architecture

agent-mcp/
├── mcp-server/     # MCP protocol server
├── agents/         # Specialized AI agents
├── memory/         # Shared state storage
└── config/         # Configuration

Agents

MCP Tools

• list_files - List files in directories

• get_file_content - Read file contents

• create_task - Create tasks in shared queue

• update_task - Update task status

• send_message - Agent-to-agent communication

• get_messages - Retrieve messages

• read_project_context - Analyze project structure

• review_code - Static code analysis

Quick Start

# Install dependencies
pip install -r requirements.txt

# Run MCP server
python -m mcp_server.server

# Or with Docker
docker build -t agent-mcp .
docker run -p 8080:8080 agent-mcp

Integration

OpenCode MCP Config

{
  "mcpServers": {
    "agent-team": {
      "command": "python",
      "args": ["-m", "mcp_server.server"],
      "cwd": "/path/to/agent-mcp"
    }
  }
}

Kilo AI Connection

The MCP server exposes a stdio interface compatible with Kilo AI's MCP client for agent communication.