Your AI coding assistant can help you add a cache in minutes. But is it reusing connections? Do the keys expire when they should? And what needs to change when that cache becomes part of an AI application?
The Azure Managed Redis agent kit is now in public preview! It puts guidance for Azure Managed Redis into your AI coding workflow to help your assistant review the choices behind code, from connection reuse to cache expiration and deployment configuration.
Whether you're adding Redis to an application, exploring semantic caching, or reviewing an Azure deployment, you can start with the project you're already working on and ask a question in plain English.
Redis guidance, right where you code
The agent kit packages instructions and reference material as Agent Skills that compatible coding assistants can load for relevant tasks. You use it in your existing coding assistant, alongside the project you're building or reviewing.
The preview includes two skills for application code and infrastructure.
- redis-code helps with writing, reviewing, and refactoring Redis application code, from client connections and caching to AI application patterns.
- azure-redis-infra helps with provisioning and configuring Azure Managed Redis, including infrastructure as code and secure, reliable deployments.
Each skill organizes its guidance into focused rules with explanations, examples, and links to documentation. The kit uses progressive disclosure to keep that guidance targeted. Once a skill activates, it directs your assistant to load the rules relevant to your task rather than the entire library. A connection-handling review doesn't need the full guide to vector search.
The kit gives your assistant reusable review criteria you would otherwise have to explain yourself. The rules connect familiar concerns, such as reusing clients, with details such as authentication token refresh on long-lived connections. Infrastructure guidance calls for reviewing module, cluster-policy, and networking choices before deployment. You can inspect the rules and their sources, then judge the recommendations against your project's requirements.
When to reach for the kit
Start with a specific question about a project you already have.
- Give an existing app a Redis checkup. Ask your assistant to review connection handling, retries, key design, and expiration. Look for issues before they become harder to diagnose in production.
- Build your next application feature. Work through data structures and caching patterns for sessions, counters, leaderboards, or messaging.
- Bring Redis into your AI application. Explore vector search for retrieval-augmented generation (RAG), semantic caching, and persistent conversation or agent state.
- Review the deployment behind the code. Ask for help with Bicep or Terraform, private connectivity, authentication, availability, and monitoring.
Get started
You'll need Node.js with npx available and a coding assistant that supports Agent Skills. From your project directory, run the following command.
npx skills add AzureManagedRedis/amr-agent-kit
The installer fetches the skills from GitHub without requiring you to clone the repository. Select one or both skills and choose the coding assistant you use. You can also specify the target with -a github-copilot for GitHub Copilot or -a claude-code for Claude Code.
Installation is project-local by default. Add -g if you want to use the skills across projects.
Then open your project in your coding assistant and start a fresh session. For a full review, point your assistant to the Redis client code and any Bicep or Terraform files in your project, then try this prompt.
Use redis-code and azure-redis-infra to review this project's use of Azure Managed Redis. Check application code and infrastructure for reliability, security, and performance issues.
Prioritize the findings, point to affected files, and explain the recommended changes. Identify anything you couldn't verify.
You can also start smaller. Try asking, "Review my Redis caching implementation for connection handling and expiration issues." Once your assistant has discovered the installed skills, it can activate them for relevant questions without you naming them. Naming a skill, as in the longer example, makes your intent explicit.
Check your assistant's skill-loading activity to confirm which skill was used. If it doesn't appear, confirm that you installed it for the assistant you're using and reload its skills or start a new session. Review the findings against the affected code and linked documentation. Ask for clarification where the assistant couldn't verify a setting, then choose a change to test.
You don't need an Azure subscription to install the kit or review code. Working with a live cache or deploying resources requires your own tools, authentication, and permissions. As with any AI-generated guidance, review recommendations and test changes before production use.
Help shape the preview
Try the kit on a real project and tell us what helped, what was missing, and where the guidance could be better. Report installation problems, unexpected output, and feature requests through GitHub issues. Please use sanitized examples without credentials, customer data, or confidential code.
For problems with the underlying Redis service or Azure infrastructure, continue to use your normal Azure support channels.
Explore the agent kit on GitHub and try a focused review on your own code!
Learn more
- New to Azure Managed Redis? Read the overview and create your first instance.
- Learn about our recent feature releases.