Inventory Operations Manager

Inventory Operations Manager is an AI agent that helps you manage and analyze your inventory with natural language.

Comment

Inspiration

Recently I asked a small retail shop owner in Sri Lanka what their main pain points are in managing their business. What I got to know from the conversation was that it would help him greatly if there was a way to get notified of items that are about to expire and get analysis information regarding inventory to optimize order quantities. It got me thinking, with the advancement of AI technology can I build a tool to support such small business owners. This thought process led me to develop an agentic inventory manager to support such business owners. To reduce the cost for the end user, a cloud based SaaS Multi Tenant approach was taken for the system.

What it does

The agent is capable of performing the following tasks,

  1. Perform CRUD operations on the database using MCP tools provided by the Inventory Management System MCP Server and MongoDB MCP Server. Only read operations go through MongoDB MCP, all alter operations go through the inventory system MCP for validation and security purposes. This means each action of the agent fully complies with the business logic. Any action that is done through the UI and more can be done with the help of the agent.
  2. Perform data analytics on data to give data driven insights regarding the business (Since the MongoDB MCP Server was used for read operations, it provides all the features to search through the database and transform it).
  3. Notify the owner of any expired / low stock items and make reorder suggestions.

How we built it

This agent was built to go with a multi tenant cloud application. The tenant configuration and user roles are provided as context for the agent to limit agent replies to the tenant. The tools used by the agent are also based on the user roles provided in the context. For read operations the agent connects to the MCP server provided by the database vendor. This ensures that the agent has all the required tools to run analysis on data based on user queries. But using the database MCP server for alter operations will mean non compliance for ACID principles and business logic. Therefore the agent is directed to use the tools provided by the system API for alter operations. The agent is built using strands SDK. The frontend of the application is built with React.js and the backend with Spring Boot.

Challenges I ran into

It was challenging to connect the system to the MongoDB MCP Server. It was also challenging to learn using Docker to deploy to AWS Bedrock AgentCore.

Accomplishments that I'm proud of

I'm proud that I was able to create an application that can support small business owners with inventory management.

What I learned

I learned how to integrate AI into the our systems using an agentic framework such as the Strands SDK and MCP protocol through this project.

What's next for Inventory Operations Manager

The inventory operations manager agent has room for further advancement. The features that come to my mind are;

  1. Send emails to suppliers
  2. Store chat history
  3. Use a vector storage database such as Amazon S3 Vectors for information retrieval from sources such as documentation.
  4. Include more features to prevent prompt injection attacks.

Built With

Share this project:

Updates

Submission history