This browser is no longer supported.
Upgrade to Microsoft Edge to take advantage of the latest features, security updates, and technical support.
A developer joins a team that owns an unfamiliar Azure Cosmos DB for NoSQL account. Before writing any application code, they want their AI assistant in Visual Studio Code to describe the containers and sample the documents inside them. Which tool fits, and what does the developer still have to configure?
The Agent Memory Toolkit, after installing the package and running its Azure Developer CLI template.
The Azure MCP Server, after signing in with the Azure CLI and receiving a data-plane role assignment on the account.
The Azure Cosmos DB MCP Toolkit, after deploying it to Azure Container Apps and issuing a bearer token.
An agent uses the Azure MCP Server text-search tool on an indexed description property. A CONTAINS query in Data Explorer finds a partial-word match, but the tool doesn't. What explains the difference?
CONTAINS
The server's identity holds the reader role instead of the contributor role, so text search operations are rejected.
Schema inference didn't detect the property, so the server excluded it from the search.
The tool uses tokenized full-text matching, while CONTAINS can match substrings.
A team wants its shopping assistant to recall user preferences across sessions without implementing its own memory-processing pipeline. Which tool best fits this requirement?
Store conversations, derive memories, and retrieve relevant user context with the Agent Memory Toolkit.
Inspect catalog documents and query product information through an editor with Azure MCP Server.
Search product documents and combine evidence across retrieval passes with the Agentic Retrieval Toolkit.
A product team is deciding between the Agentic Retrieval Toolkit and a single vector search for a support assistant. Most questions are direct lookups answered by one document, but roughly one in 20 spans several documents. What's the sound recommendation?
Use agentic retrieval for every question, because a consistent retrieval path is easier to operate than two paths.
Use a single query for the direct lookups and route only the multi-document questions to the iterative pipeline.
Use a single query for everything and raise the number of retrieved documents, because more context makes the multi-document questions answerable.
A team adopts the Agent Memory Toolkit and the Agentic Retrieval Toolkit, then asks which responsibilities the toolkits take over. Which statement correctly describes the boundary?
The toolkits own security, because both authenticate with Microsoft Entra ID by default and reach the account through managed identities.
The toolkits own the data model, because the memory toolkit provisions its containers with a hierarchical partition key.
The toolkits own the storage and retrieval patterns, while throughput, indexing policy, the scope of data-plane role assignments, and answer quality remain the team's responsibility.
You must answer all questions before checking your work.
Was this page helpful?
Need help with this topic?
Want to try using Ask Learn to clarify or guide you through this topic?