Introducing the Data Kiosk MCP Server: Supercharge Your Amazon Analytics with AI

[hichenai]

In the competitive world of e-commerce, data is king. Amazon sellers and vendors need deep insights into their business performance to make informed decisions. While Amazon provides rich data through their Selling Partner API, many developers and sellers are looking to enhance their analytical capabilities through custom solutions and AI integration. This creates an opportunity to demonstrate how developers can build well-architected, AI-powered tools that complement existing solutions in the marketplace.

Today, we're excited to introduce the Data Kiosk MCP Server Sample Solution – a powerful new way to interact with your Amazon business data using advanced AI assistants.

This MCP server example runs locally in your test environment and provides a foundation for development and experimentation. Developers utilizing this sample solution are responsible for its implementation, usage patterns, and ensuring compliance with Amazon's Acceptable Use Policy (AUP) and Data Protection Policy (DPP)

0729.mov

This project is primarily aimed at:

1. Helping business users access Data Kiosk data through natural language without needing to understand GraphQL or set up data infrastructure
2. Enabling developers to quickly prototype and experiment with GraphQL queries
3. Serving as a demonstration tool rather than a production solution

AI Compliance

In accordance with Amazon's SP-API policies, this implementation strictly prohibits the use of Amazon or seller production data for training Data Analytics and AI models. All data access is managed securely through official SP-API endpoints using your Selling Partner API credentials. When implementing AI features, you must use self-hosted models or ensure your AI service providers have contractual agreements preventing the use of prompt data for training purposes. Each seller's data must remain isolated - the API outputs cannot be processed in aggregate form or shared across different sellers' accounts to maintain data privacy and prevent cross-seller data exposure. For complete information about our AI compliance policies and requirements, please visit the official DPP and AUP documentation. By using this tool, you agree to comply with all Amazon Selling Partner API terms, conditions, and policies, including specific requirements regarding AI and data analytics usage.

What is MCP and Why Should You Care?

The Model Context Protocol (MCP) is an open protocol that allows AI assistants to access external tools and data sources. By connecting Large Language Model (LLM) to your Amazon data through our MCP server, you can have natural conversations with your business data – asking questions, generating reports, and uncovering insights without writing a single line of code. Learn more about MCP

Unlike traditional dashboards or analytics platforms, the Data Kiosk MCP server lets you:

• Ask questions in plain English about your sales, traffic, and financial performance
• Generate custom reports and visualizations tailored to your specific needs
• Combine data across different Amazon domains (sales, traffic, economics, etc.)
• Get actionable recommendations based on your business trends

How the Data Kiosk MCP Server Works

The diagram above illustrates how the Data Kiosk MCP server bridges the gap between AI and your Amazon business data:

1. You start a conversation with an AI assistant in an MCP-enabled client, asking questions about your Amazon business. Find MCP clients

2. The MCP Client uses the MCP protocol to determine which tools are needed to answer your question

3. The Data Kiosk MCP server receives tool calls from the MCP Client and:

   • Uses common utilities for authentication and API handling
   • Leverages specialized servers for Seller or Vendor analytics
   • Employs domain-specific tools for different types of data
   • Constructs appropriate GraphQL queries through the Query Builder

4. The Amazon Data Kiosk API receives these queries and returns the relevant business data

5. The LLM processes and analyzes the returned data to provide you with insights in natural language

This architecture allows you to interact with complex Amazon data without needing to understand GraphQL, API authentication, or data processing techniques.

What Data Can You Access?

The Data Kiosk MCP server provides comprehensive access to your business data across multiple domains:

For Amazon Sellers:

1. Sales and Traffic Data

   • Daily, weekly, or monthly sales metrics
   • Traffic and conversion statistics
   • Product performance by ASIN, parent ASIN, or SKU
   • B2B vs. general marketplace metrics

2. Economics Data

   • Detailed fee breakdowns across all Amazon fee types
   • Profitability analysis by product
   • Cost impact analysis
   • Future fee change impact previews

For Amazon Vendors:

3. Vendor Analytics

   • Sourcing metrics for distributor relationships
   • Manufacturing metrics for your own products
   • Inventory performance and forecasting
   • Shipment and order metrics

We are committed to keeping this project up-to-date with Amazon's evolving Data Kiosk API. As Amazon adds new metrics, data domains, or enhanced capabilities to their Data Kiosk service, we will regularly update the MCP server to incorporate these new features, ensuring you always have access to the latest analytics capabilities through MCP.

How User Interaction Works

The Data Kiosk MCP server acts as a bridge between AI assistants and the Amazon Selling Partner API:

1. You connect the server to an MCP-enabled client on your Mac or Windows computer
2. Authenticate with your Amazon Selling Partner API credentials
3. Start asking questions about your Amazon business

The Data Kiosk MCP server transforms how you interact with your Amazon business data through a seamless, conversation-based workflow.

User Interaction Workflow

The diagram above shows the streamlined workflow:

1. User Question: You ask LLM a question about your Amazon business in natural language, such as "How were my sales last month?"
2. LLM Selects Tool: LLM analyzes your question and selects the appropriate MCP tool from the available options (sales and traffic, economics, or vendor analytics).
3. Parameter Gathering: LLM collects necessary parameters for the query, such as date ranges and marketplace identifiers, either from your question or by asking follow-up questions.
4. Query Execution: The MCP server builds and submits a GraphQL query to the Amazon Data Kiosk API, translating your natural language request into the required technical format.
5. Amazon API: The Data Kiosk API processes your request, retrieving the requested data from Amazon's systems and returning it to the MCP server.
6. Results Interpretation: LLM analyzes the raw data returned from Amazon, identifying patterns, trends, and noteworthy insights.
7. Presentation: LLM presents the results in a user-friendly format with visualizations, explanations, and actionable recommendations.
8. Follow-up Analysis: You can ask additional questions to explore the data further, with LLM maintaining context from the previous interactions.

This simple workflow handles all the technical complexity – authentication, query construction, data processing – behind the scenes, letting you focus on the business insights that matter most. You can continue the conversation with follow-up questions, and the system will repeat this process, maintaining context from your previous interactions.

Hands-On Getting Started

Setting up the Data Kiosk MCP server involves a few simple steps:

1. Clone the repository from GitHub:

git clone https://github.com/amzn/selling-partner-api-samples.gitcd use-cases/amazon-data-kiosk-mcp-server

1. Install dependencies and build the project:

npm install
npm run build

1. Configure MCP Client to use the server:

Locate your MCP server configuration file. Then add the server configuration:
Note: While you can add both the seller and vendor servers to your configuration, it's recommended to add only one at a time to avoid confusion for the AI model. Choose the server that matches your account type (seller or vendor).

{
    "mcpServers": {
        "amazon-seller-analytics": {
                "command": "node",
                "args": [
                "/absolute/path/to/amazon-data-kiosk-mcp-server/packages/seller-server/build/index.js"
                ],
                "env": {
                "DATA_KIOSK_CLIENT_ID": "your_client_id",
                "DATA_KIOSK_CLIENT_SECRET": "your_client_secret",
                "DATA_KIOSK_REFRESH_TOKEN": "your_refresh_token",
                "DATA_KIOSK_BASE_URL": "https://sellingpartnerapi-na.amazon.com",
                "DATA_KIOSK_OAUTH_URL": "https://api.amazon.com/auth/o2/token",
                "DATA_KIOSK_API_VERSION": "2023-11-15"
            }
        }
    }
}

Make sure to replace the paths with the actual absolute paths to the built JavaScript files on your system.

Example Conversations

Here are a few examples of how you can interact with your Amazon data through an AI assistant:

Example 1: Visualizing Sales by Day and MSKU

You: "Give me sales by days and MSKU. Output in a diagram"
LLM will:

1. Create a GraphQL query targeting the sales data
2. Submit it to the Amazon Data Kiosk API
3. Check the query status until complete
4. Download and process the results
5. Generate an interactive visualization showing daily sales by MSKU
6. Provide insights about the data patterns

As shown in the example, LLM can identify key patterns such as:

• Which products consistently show the highest sales volume
• Weekend sales spikes across products
• Overall sales trends and growth patterns
• Relative volatility between different products

This visualization capability makes it easy to spot patterns that might be missed when looking at raw data tables.

Example 2: Fee Impact Analysis

You: "How will the FBA fee changes coming next month affect my profitability?"
AI: [Uses MCP server to run fee preview queries and analyze results]
"I've analyzed the upcoming FBA fee changes effective June 1, 2025, for your catalog:
Overall impact: Your average profit margin will decrease by approximately 1.2% across all products.

Most affected categories:

• Oversize items (>3 lbs): Average fee increase of $1.43 per unit
• Apparel: Average fee increase of $0.78 per unit
• Small standard size: Minimal impact (avg. $0.12 increase)

Top 3 products requiring attention:

1. Camping Tent (B09X123Y45Z): Profit margin will decrease from 24% to 19%
2. Winter Jacket (B07A987C654): Profit margin will decrease from 32% to 28%
3. Desk Lamp (B09D321E456): Profit margin will decrease from 18% to 16%

Would you like me to recommend pricing adjustments to maintain your target margins?“

Example 3: Inventory Optimization

You: "Analyze my inventory health and identify potential issues."
AI: [Uses MCP server to query inventory metrics and perform analysis]
"Based on your current inventory data, here's my analysis:
Overall inventory health score: 82/100 (Good, but with room for improvement)

Key findings:

• 72% of your inventory is selling within expected timeframes
• 18% of inventory (23 ASINs) is aging past 90 days, representing $34,562 in capital
• 5 products are at risk of stockout within the next 7 days
• Your average sell-through rate is 18% weekly (healthy for your category)

Recommendations:

1. Consider running promotions for these 5 slow-moving ASINs: [list]
2. Restock these 3 fast-selling items immediately: [list]
3. Review your reorder quantity for seasonal items - you're currently overstocked on summer products by 34%

Would you like me to help create a restock plan or promotion strategy for any of these items?“

Technical Architecture

The Data Kiosk MCP server is built with a modern, modular architecture:

• Node.js backend for efficient processing and API communication
• GraphQL query builder to translate natural language requests to Amazon's API format
• Authentication management to handle secure API connections
• Data processing utilities to transform raw Amazon data into actionable insights

The codebase is organized into three main packages:

• common: Shared utilities and base functionality
• seller-server: MCP server for Amazon Seller Analytics
• vendor-server: MCP server for Amazon Vendor Analytics

Get Started Today

The Data Kiosk MCP server is hosted on GitHub and runs locally on your machine, ensuring your Amazon business data remains private and secure. Whether you're a small seller looking to grow your business or an established vendor managing complex operations, this tool can help you unlock the power of AI-assisted analytics for your Amazon business.

To get started:

1. Make sure you have an MCP-enabled client installed
2. Get your Amazon Selling Partner API credentials ready
3. Clone the GitHub repository and follow the setup instructions
4. Start analyzing your Amazon data with AI!

Join our community to share feedback, report issues, or contribute to the project's development.
Transform your Amazon business intelligence today with the power of Large Language Models and the Data Kiosk MCP server!

Authored by: @hichenai Solutions Architect, Selling Partner Developer Services

Let us know what do you think! 👇 React or reply down below to share your feedback. Happy shopping!

• ❤️ Like what you see?

• 👍 Excited to try it out?

• 🎉 Want to see more of this?

Keywords: amazon selling partner api, data kiosk, mcp, model context protocol, LLM, ai analytics, graphql, amazon seller, amazon vendor, business intelligence, sales dashboard, mcp client, amazon api integration

Please note the publication date above. While we strive to keep our content current, the rapid evolution of this topic means some details may have changed since publication. For the most up-to-date information, we encourage readers to follow the announcements in SP-API developer documentation.

Your vote takes 1 sec but it will help us create more of what you love! 💬