Build an Agentic AI Assistant with Strands Agents SDK and AWS MCP Servers

Introduction

Strands Agents is an open-source SDK that simplifies the development of AI agents capable of using tools, making decisions, and automating workflows, moving far beyond fundamental chatbot interactions.

MCP (Model Context Protocol) is an open protocol that enables LLM-powered agents to connect to external tools and data sources securely. AWS provides official MCP servers for both documentation and pricing , making it possible for an agent to ground its responses in live, authoritative data.

The stdio transport runs MCP servers locally on your machine via standard input/output streams. It’s simple and secure for development, while Streamable HTTP is better suited for production or distributed environments.

In modern cloud projects, solution architects often face questions like.

What’s the difference between Amazon Kendra Enterprise Edition and the Generative AI Index?
Which is better for integrating SharePoint Online data sources, Amazon Bedrock Knowledge Base, or Kendra?
What will it cost to run 8 million documents with ACL passthrough enabled?

Answering these requires real-time access to both.

Features and capabilities
Pricing and cost estimates

Manually digging into docs and calculators can slow things down. This is where Agentic AI comes in. Using the Strands Agents SDK together with AWS MCP servers, we can create an intelligent agent that.

AWS Documentation MCP server: Retrieves official AWS documentation for features, limitations, and capabilities
AWS Pricing MCP Server: Pulls real-time pricing data from the AWS Pricing API
Generates comparisons with pros, cons, and recommendations
(Optional) Provides outputs in customer-facing Markdown tables ready to share

In this article, you will learn the following.

How to integrate multiple AWS MCP servers into one agent
Using the Strands Agents SDK to orchestrate documentation + pricing queries
How to build a Solution Architect Assistant Agent that outputs.
- Feature comparison tables
- Pros and cons
- Pricing estimates with real-time data
- A scenario-driven recommendation

Pre-requisites

Install or update to the latest version of the AWS CLI.
Get credentials to grant programmatic access.
Visual Studio Code.
Access to the Amazon Bedrock foundation model. The default model provider is Amazon Bedrock, and the default model is Claude 3.7 Sonnet in the US Oregon (us-west-2) region.

Create an agent using Strands Agents SDK

Perform the following steps to create and configure an agent using Strands Agents SDK.

Open Visual Studio Code.

Navigate to the folder where you want to create your Python file.

Open a new PowerShell terminal in Visual Studio Code.

Run the following command to create a virtual environment to install the Strands Agents SDK and its dependencies.

  
    python -m venv .venv

Run the following command to activate the virtual environment.

  
    .venv\Scripts\Activate.ps1

Run the following command to install the strands-agents SDK package.

  
    pip install uv
pip install strands-agents strands-agents-tools mcp

Create a new Python file and name it feature_comparison_agent.py.

Copy and paste the code below into feature_comparison_agent.py.

  
    import os
from mcp import StdioServerParameters, stdio_client
from strands import Agent
from strands.tools.mcp import MCPClient

# ---- Set these in PowerShell before running (or edit defaults here) ----
AWS_PROFILE = os.environ.get("AWS_PROFILE", "default")
AWS_REGION  = os.environ.get("AWS_REGION",  "us-east-1")
LOG_LEVEL   = os.environ.get("FASTMCP_LOG_LEVEL", "ERROR")

def docs_stdio_params() -> StdioServerParameters:
    return StdioServerParameters(
        command="uvx",
        args=[
            "--from", "awslabs.aws-documentation-mcp-server@latest",
            "awslabs.aws-documentation-mcp-server.exe",
        ],
        env={"FASTMCP_LOG_LEVEL": LOG_LEVEL},
    )

def pricing_stdio_params() -> StdioServerParameters:
    return StdioServerParameters(
        command="uvx",
        args=[
            "--from", "awslabs.aws-pricing-mcp-server@latest",
            "awslabs.aws-pricing-mcp-server.exe",
        ],
        env={
            "FASTMCP_LOG_LEVEL": LOG_LEVEL,
            "AWS_PROFILE": AWS_PROFILE,
            "AWS_REGION": AWS_REGION,
        },
    )

# Build MCP clients (stdio transport)
aws_docs_mcp_client    = MCPClient(lambda: stdio_client(docs_stdio_params()))
aws_pricing_mcp_client = MCPClient(lambda: stdio_client(pricing_stdio_params()))

# System Prompt
system_prompt = """
You are an AWS Solution Architect Assistant.

Behavior:
- For FEATURES / CAPABILITIES / QUOTAS / KNOWN ISSUES OR LIMITATIONS: Prefer AWS Documentation MCP tools.
- For PRICING: Prefer AWS Pricing MCP tools for live prices (don’t guess).
- Present results in clean Markdown with three sections:
  1) **Feature Comparison** — a table
  2) **Pros & Cons** — bullet list per option
  3) **Pricing** — region + assumptions, clearly stated
- End with **Recommendation** — concise and scenario-driven.

Citations:
- After each feature/limit claim sourced from docs, include a short “Source:” line
  referencing the doc title or URL provided by the AWS Documentation MCP.
- For pricing, note “Source: AWS Pricing MCP (real-time)”.

Formatting:
- Use concise, scannable language suitable for customer-facing notes.
- If a product/edition cannot meet a stated requirement (e.g., quotas), say so explicitly.
"""

# User Prompt
prompt = """
Compare **Amazon Kendra Gen AI Index** vs **Amazon Kendra Enterprise Edition**
for building a SharePoint Online knowledge base.

Include:
- A feature table focusing on: supported auth modes for SharePoint Online, connector GA vs preview status, ACL passthrough, scale/quotas, ingestion options, RAG/LLM integration path.
- Pros & cons for each.
- Pricing estimate in **us-east-1** for a hypothetical workload:
  - ~8 million documents
  - ~12 QPS for query traffic
  - Call out any quota blockers explicitly.
- Finish with a clear recommendation by scenario (e.g., enterprise compliance, rapid PoC, heavy SharePoint ACLs, very large corpora).
Make sure you use documentation tools for facts and the pricing tool for cost.
"""

if __name__ == "__main__":
    # Keep tool usage INSIDE the MCP client context so sessions stay open
    with aws_docs_mcp_client, aws_pricing_mcp_client:
        tools = (
            aws_docs_mcp_client.list_tools_sync()
            + aws_pricing_mcp_client.list_tools_sync()
        )

        agent = Agent(
            system_prompt=system_prompt,
            tools=tools,
            model="us.anthropic.claude-3-7-sonnet-20250219-v1:0",  # Bedrock Claude 3.7 Sonnet
        )

        print("\n[🔍 Asking Agent]:\n", prompt)
        response = agent(prompt)
        print("\n[📊 Agent Response]:\n")
        print(response)

Run the following command to execute your Python code and to view the results.

  
    python.exe .\feature_comparison_agent.py

Summary

In this article, you learned how to build an agentic AI assistant that integrates with multiple AWS MCP servers using the Strands Agents SDK. By combining AWS Documentation MCP for features and limitations with AWS Pricing MCP for live cost data, your agent can generate rich, customer-ready comparisons complete with pros/cons, tables, and recommendations. This approach demonstrates how Agentic AI can expedite solution architecture decisions and minimize manual effort.