Azure Cosmos DB Agent Kit overview

The Azure Cosmos DB Agent Kit is a collection of Agent Skills that gives AI coding assistants Azure Cosmos DB guidance for data modeling, partition keys, queries, and SDK usage. Use the kit when you build or review an application and want recommendations for performance, scalability, and cost.

The kit is designed for assistants that load relevant skills on demand, rather than adding every rule to every prompt. You can find the skills and installation options in the Agent Kit repository.

Quick start

To get started, you need an AI coding assistant that supports Agent Skills and Node.js with npm and npx installed.

1. Install the kit:

   npx skills add AzureCosmosDB/cosmosdb-agent-kit
   

2. Follow the installation prompts for your coding assistant. For other installation methods, see Set up the Agent Kit.

3. In your assistant, provide your data model and workload requirements, and try this prompt:

   Use the Azure Cosmos DB Agent Kit to review my proposed partition key.
   Consider my query patterns, write distribution, and expected data growth.
   Explain the trade-offs and identify any missing information before recommending a key.
   

Review the response for recommendations tied to the workload information you provided. Check your host's skill-loading indicators, if available, to confirm that it loaded the relevant skill. For more task examples, see Use the Agent Kit.

What the kit helps with

The kit focuses on the decisions that affect performance, scalability, and cost:

• Partition key selection.
• Query optimization.
• Data modeling.
• Hot partition prevention.
• SDK usage patterns.
• Security and access patterns.
• Indexing strategy.
• Throughput and scaling.
• Global distribution.
• Developer tooling.
• Monitoring and diagnostics.
• Vector search.
• Full-text search.
• Design patterns.

Agent skills and MCP servers

Agent skills provide guidance. Installing the Agent Kit doesn't, by itself, authenticate your assistant to an Azure account or give it permission to query a database. Treat generated recommendations as suggestions to review and test, not as evidence of measured performance improvements.

The Microsoft Learn MCP server provides a separate capability: it lets compatible assistants search official documentation and retrieve articles. It doesn't provide access to your Azure Cosmos DB data. Database operations require separately configured tools and permissions.

For host-specific considerations, see Supported AI coding assistants.

AI tools

Connect your coding agent to Microsoft Learn or read this article in your assistant.

For a compact index of the guides and their Markdown versions, use the Agent Kit documentation index (llms.txt). This index is a discovery aid, not a crawler policy or a guarantee that an assistant retrieves the articles.

Connect your AI agent
Set up Microsoft Learn MCP in Visual Studio Code.

Read as Markdown
Open a text version to copy into your assistant.

Ask ChatGPT
Start a conversation about this article in ChatGPT.

Ask Claude
Start a conversation about this article in Claude.

ChatGPT and Claude open external services with the public article URL and a starter prompt. You might need to sign in. If the assistant can't access the article, copy the Markdown text instead. These actions don't install the kit or grant database access.

Watch the overview

Watch a short introduction to the Agent Kit.

Next step