A developer reviews an Azure Managed Redis project with an AI coding assistant and cloud architecture diagram.
Microsoft has put an Azure Managed Redis agent kit into public preview, and it's aimed at the AI coding assistant on your developer machine, not at your cache. The goal is to make your assistant review Redis code and Azure deployments with real criteria, not a generic "looks fine."

The announcement says the kit is now in public preview. It puts guidance for Azure Managed Redis into your AI coding workflow so your assistant can review the choices behind code, from connection reuse to cache expiration and deployment configuration. The Azure Updates entry gives the item an "In preview" status. That label means it is available for non-production use and testing, not production-ready.

A developer reviews an Azure Managed Redis project with an AI coding assistant and cloud architecture diagram. What the kit actually is​

The kit is not an agent runtime. Microsoft's Ignite roundup says the kit supports development and operational decisions and is not an agent runtime. It's a set of instructions and reference material, packaged as Agent Skills that compatible coding assistants can load for relevant tasks.

The preview ships two skills:

  • redis-code helps write, review and refactor Redis application code. That covers client connections, caching and AI application patterns.
  • azure-redis-infra helps provision and configure Azure Managed Redis. That includes infrastructure as code and secure, reliable deployments.

Each skill breaks its guidance into focused rules, with explanations, examples and documentation links. Microsoft calls the approach "progressive disclosure." Once a skill activates, it points the assistant at the rules relevant to the task, not the whole library. A connection-handling review doesn't need the vector-search guide.

What it checks​

Microsoft's examples show the intended depth:

  • Reusing clients is a familiar concern. The kit ties it to less obvious details, such as authentication token refresh on long-lived connections.
  • Infrastructure guidance asks for review of module, cluster-policy and networking choices before deployment.
  • Microsoft's Ignite Book of News says the kit will identify configuration settings that could affect reliability, network connections and performance, then suggest changes for developers to evaluate.
  • The Azure Updates summary, as mirrored by Cloud Changelog, also mentions guided workflows for migrating from Azure Cache for Redis and Azure Cache for Redis Enterprise to Azure Managed Redis.

Suggested uses from Microsoft's post:

  1. Give an existing app a Redis checkup covering connection handling, retries, key design and expiration.
  2. Work through patterns for sessions, counters, leaderboards or messaging.
  3. Explore vector search for RAG, semantic caching, and persistent conversation or agent state.
  4. Review Bicep or Terraform, private connectivity, authentication, availability and monitoring.

These are tasks the kit can help an assistant look at. Microsoft doesn't promise it will catch every issue or validate a deployment.

How to install it​

You need Node.js with npx and a coding assistant that supports Agent Skills. From your project directory, run:

npx skills add AzureManagedRedis/amr-agent-kit

The installer fetches the skills from GitHub without requiring you to clone the repository. Then:

  1. Select one or both skills and pick your assistant. You can target one directly with -a github-copilot or -a claude-code.
  2. Installation is project-local by default. Add -g to use the skills across projects.
  3. Open the project in your assistant and start a fresh session.
  4. For a full review, point the assistant at your Redis client code and any Bicep or Terraform files.

Microsoft's sample prompt asks the assistant to use both skills to review the project's use of Azure Managed Redis. It covers reliability, security and performance. It also asks the assistant to prioritize findings, point to affected files, explain the recommended changes, and say what it couldn't verify. A smaller prompt works too, for example a review of caching connection handling and expiration. Once the assistant has discovered the installed skills, it can activate them without being named. Naming them makes your intent explicit.

If it doesn't seem to work​

  • Check your assistant's skill-loading activity to see which skill was used.
  • If the skill doesn't appear, confirm you installed it for the assistant you're actually using.
  • Then reload skills or start a new session.
  • Check findings against the affected code and the linked documentation. Where the assistant couldn't verify a setting, ask for clarification, then pick a change to test.

Prerequisites and limits​

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. Microsoft also warns that, as with any AI-generated guidance, you should review recommendations and test changes before production use.

For feedback, Microsoft points to GitHub issues for installation problems, unexpected output and feature requests. It asks for sanitized examples with no credentials, customer data or confidential code. Problems with the Redis service or Azure infrastructure should go through normal Azure support channels.

Why it matters​

Most AI-generated cache code works in the demo. The risk shows up later: connections that aren't reused, keys that never expire, tokens that lapse on long-lived connections. A reusable rule set gives teams a consistent baseline, so they don't have to re-explain Redis best practices in every prompt. Because the rules and their sources are inspectable, you can judge them against your own requirements.

There are caveats. This is guidance fed to a language model, so output quality still depends on the assistant and on your project. Nothing here replaces code review or load testing. It also doesn't change what the service itself does.

Where it sits in the Ignite announcements​

Microsoft's Ignite 2026 roundup lists seven Azure Managed Redis updates on November 17, 2026. The agent kit is one of two lead previews, alongside Redis 8. Shard-level metrics and IP filtering are also previews. Migration tooling, database integration guidance and keyspace notifications are generally available. Those are separate announcements with their own status, so don't read the kit's preview label onto them, or the reverse.

If you're evaluating Azure Managed Redis, the sensible approach is to run the kit on a non-critical branch first. Compare its findings with your own review, and note what it misses. Then send that feedback to Microsoft while the preview is still open.

 

References

  1. (In preview) Public Preview: Azure Managed Redis agent kit Azure Updates 2026-11-17T20:00:32Z
  2. Go beyond “it works” and build better apps with the Azure Managed Redis agent kit techcommunity.microsoft.com
  3. What’s new in Azure Managed Redis at Microsoft Ignite 2026 techcommunity.microsoft.com