Building an Agentic Warehouse Management System – Part 3: Tools, Decisions, and Actions
MongoDB, Thursday, November 5th, 2026
The last part of MongoDB's agentic WMS series shows a Java and Spring AI agent executing tasks with controlled tools to make replenishment decisions.
Part 3 of MongoDB's agentic warehouse management system series covers the execute, decide and complete stages of the agent workflow built with Java and Spring AI.
An AgentRunner processes planned tasks one at a time, giving each task only the tools tied to its capability, so ANALYSIS tasks can read WMS data such as inventory and stock movements but cannot create replenishment requests.
POLICY tasks use a simple RAG flow, retrieving depositor policies stored as natural-language documents and embedded with a Voyage AI model in a vector store.
A DECISION task then determines whether replenishment is needed; if not, remaining tasks are skipped, and if so, the agent creates a replenishment request through application services and drafts a notification email.