Strands Agent MCP

A Model Context Protocol (MCP) server for executing Strands agents. This project provides a simple way to integrate Strands agents with Amazon Q and other MCP-compatible systems.

IMPORTANT: This project is currently in alpha stage and not yet published on PyPI.

Overview

Strands Agent MCP is a bridge between the Strands agent framework and the Model Context Protocol (MCP). It allows you to:

• Register Strands agents as MCP tools

• Execute Strands agents through MCP

• Find agents by specific skills

The project uses a plugin architecture that makes it easy to add new agents without modifying the core code.

Related MCP server: A2A Client MCP Server

Installation

Note: This package is not yet available on PyPI. You'll need to install it from source.

# Clone the repository
git clone https://github.com/yourusername/strands-agent-mcp.git
cd strands-agent-mcp

# Install the package
pip install -e .

Usage

Starting the MCP Server

strands-agent-mcp

This will start the MCP server.

Environment Variables

The server supports the following environment variables:

• PLUGIN_PATH: Custom path to look for plugins (default: ".")

• PLUGIN_NAMESPACE: Custom namespace prefix for plugins (default: 'sap_mcp_plugin')

Creating Agent Plugins

To create a new agent plugin, create a Python package with a name that starts with sap_mcp_plugin_ (sap stands for strands agent plugin). Your package should implement a build_agents function that returns a list of AgentEntry objects:

from typing import List
from boto3 import Session
from strands import Agent
from strands.models import BedrockModel

from strands_agent_mcp.registry import AgentEntry

def build_agents() -> List[AgentEntry]:
    return [
        AgentEntry(
            name="my-agent",
            agent=Agent(
                model=BedrockModel(boto_session=Session(region_name="us-west-2"))
            ),
            skills=["general-knowledge", "coding"]
        )
    ]

Using with Amazon Q

Once the MCP server is running, you can connect it to Amazon Q. Refer to the Amazon Q documentation for the correct connection parameters.

The following MCP tools will be available:

• execute_agent: Execute an agent with parameters agent_name and prompt

• list_agents: List all available agents

Architecture

The project consists of three main components:

1. Server: The MCP server that exposes the agent execution API

2. Registry: A registry for managing available agents and their skills

3. Plugins: Dynamically discovered modules that register agents with the registry

The server automatically discovers all installed plugins that follow the naming convention and registers their agents.

Dependencies

• fastmcp>=2.3.4: For implementing the MCP server

• strands-agents>=0.1.1: The core Strands agent framework

• strands-agents-builder>=0.1.0: Tools for building Strands agents

• strands-agents-tools>=0.1.0: Additional tools for Strands agents

Development

This project uses uv for dependency management. To set up a development environment:

1. Clone the repository

2. Install uv if you don't have it already: pip install uv

3. Create a virtual environment and install dependencies:

   uv venv
   uv sync

Sample Plugin

The repository includes a sample plugin (sap_mcp_plugin_simple) that demonstrates how to create and register a simple agent:

from typing import List
from boto3 import Session
from strands import Agent
from strands.models import BedrockModel

from strands_agent_mcp.registry import AgentEntry

def build_agents() -> List[AgentEntry]:
    return [
        AgentEntry(
            name="simple-agent",
            agent=Agent(
                model=BedrockModel(boto_session=Session(region_name="us-west-2"))
            ),
            skills=["general-knowledge"]
        )
    ]

License

This project is licensed under the terms of the LICENSE file included in the repository.