All things Azure
Developer focused how-tos, use cases and solutions on Microsoft Azure
Featured posts
Azure Skills Plugin – Let’s Get Started!
Part 2 of the Azure Skills Plugin series Previously: Announcing the Azure Skills Plugin This post is all about getting you up and running. I won't g...
Platform Engineering for the Agentic AI era
For the last decade, platform engineering has relied on explicit API interaction layers: CLIs, SDKs, pipelines, wrappers, and UI workflows that translate human ...
Context-Driven Development: Agent Skills for Microsoft Foundry and Azure
Code will be generated, not written. Most enterprise AI workloads are net-new microservices. Modular, greenfield work. Perfect for coding agents. The catch? ...
Claude Code + Microsoft Foundry: Enterprise AI Coding Agent Setup
This guide covers setting up Claude Code CLI and VS Code extension with Microsoft Foundry, configuring CLAUDE.md for project context, integrating Spec Kit for s...
Visualizing GitHub Audit Log in Microsoft Defender
Key Observability Trends Around GitHub Security Modern enterprises are increasingly adopting DevSecOps practices, integrating security into every phase...
Codex Azure OpenAI Integration: Fast & Secure Code Development
Introduction You can now enjoy the same Codex experience in CLI or VS Code with Azure OpenAI support. We've contributed the following five pull requests to mak...
How to develop AI Apps and Agents in Azure – A Visual Guide
As organizations explore new AI-powered experiences and automated workflows, there's a growing need to move beyond experiments and proofs-of-concept to producti...
Latest posts
I Wasted 68 Minutes a Day Re-Explaining My Code. Then I Built auto-memory.
~1,900 lines of Python. Zero dependencies. Saves you an hour a day. GitHub → · Now give Copilot CLI enhanced context recall. Point it at and let it cook. 🍳 Are you tired of using the slash /compact command every 10 min? The Context Window Is a Lie Every AI coding agent ships with a big number on the box. 200K tokens. Sounds massive. You could fit an entire codebase in there, right? Here's what actually happens when you start a session: 200,000 tokens — your context window (on paper) -65,000 tokens — MCP tools load at startup (~33%) -10,000 tokens — instruction files ((~5%) ========= ~12...
Getting Started with Agentic DevOps – Part 1: Foundations
This post is the first in a 3-part series: Bookmark this post for quick reference as you start exploring Agentic DevOps. It will be updated as the 3 parts become available. Getting started with Agentic DevOps Agentic DevOps is a new approach to software development where AI-powered agents work alongside your team across the entire software development lifecycle. Unlike traditional AI assistance, these agents go beyond suggestions—they can take on tasks end-to-end, collaborate across tools, and operate across the lifecycle with your guidance and approval. This series is designed as a pract...
Best of Both Worlds for Agentic Refactoring: GitHub Copilot + MicroVMs via Docker Sandbox
Legacy codebases frequently contain hardcoded logic and complex build scripts that depend on specific filesystem structures, making them notoriously difficult to modernize in isolated environments. Docker Sandbox addresses this challenge through a bidirectional workspace sync that preserves the same absolute paths inside the sandbox as on the host. This means that when a GitHub Copilot agent refactors a legacy Java or .NET application, file references and build outputs remain consistent across the isolation boundary. The result? Modernized code can be moved back to the host without breaking dependencies. Ho...
Choosing the Right Azure Hosting Model for AI Agents: A Deep Dive into Foundry Hosted Agents
AI agents are quickly moving from experiments to production‑critical components of modern applications. But while many teams know how to build agents, far fewer are confident they’re hosting them on the right foundation. Most organizations start by deploying agents the same way they deploy microservices—containers, functions, or app services. That approach works initially. But as agents evolve to support long‑running conversations, tool orchestration, stateful workflows, and continuous iteration, infrastructure decisions start to matter in new ways. Azure offers multiple ways to host AI agents, each wit...
DevOps Playbook for the Agentic Era
Practices, Principles, and Strategic Direction Software delivery has entered a new phase. AI agents are no longer confined to autocomplete suggestions in the editor. They are opening pull requests, generating code across multiple files, proposing infrastructure changes, responding to issues with working implementations, and executing multi-step engineering tasks with minimal human intervention. Tools like GitHub Copilot Cloud Agent (coding agent) represent the leading edge of a shift that is transforming how teams design, build, test, and ship software. This is not a future scenario. It is happening now, acro...
Putting Agentic Platform Engineering to the test
In Part 1 of this blog series set the stage for why platform engineering is being reshaped by agentic AI. (read it here) Basically we outline how instead of humans translating intent through layers of CLIs, SDKs, and bespoke workflows, capable agents can interpret natural-language goals and turn them into safe, validated platform actions using well-described APIs and control schemas. That shift changes what “good” looks like for internal platforms, raising the bar on guardrails, policy, and self-service interfaces allowing teams to move faster without sacrificing safety, reliability, or governance. In th...
Project Nighthawk: A Research Agent Built for Field Engineering
If you work in field engineering, you know the scenario. A customer is deploying AKS in a regulated environment. They hit an issue during node bootstrapping. They want to know exactly what happens when a node joins the cluster, which components run in which order, and whether the behaviour they're seeing is expected. The question sounds simple. The answer is not. The answer is spread across half a dozen places at once. It's in the source code: AgentBaker, the node controller, cloud-provider-azure. It's in a Microsoft Learn article that's technically correct but three levels of abstraction above what actually r...
Building with Azure Skills
Part 3 of the Azure Skills Plugin series Previously: How to Install the Azure Plugin You've installed the Azure Skills Plugin. The Azure MCP server is running. You have a huge collection of tools and skills at your disposal. So, now what do you actually say to it? This post is a prompt cookbook. Every example below is a real prompt you can type into Copilot Chat (or Copilot CLI, or Claude Code) with the Azure Plugin installed. Each one triggers a specific skill and produces a concrete, actionable result - not generic advice. A note on scope: Azure is huge, and the examples below reflect th...
Your Entire Engineering Floor Just Stopped Coding
And the developers running Claude Code and GitHub Copilot CLI didn't notice... Status Page Says 'Operational.' Your Subagents Say Otherwise. If you're running autonomous agents in any serious capacity, you've experienced this: model provider outages aren't edge cases — they're part of the operating environment. Anthropic has had outages. OpenAI has had outages. Google has had outages. Every major model provider has had the kind of degraded performance that doesn't trigger a status page alert but absolutely kills an agentic coding session. The traditional answer is "wait it out." But if you're a solution...
Agentic Platform Engineering with GitHub Copilot
We've talked about the human scale problem and what happens when infrastructure scales but understanding doesn't. If you've been following along, you know the thesis: our tools have outpaced our ability to operate them, and platform engineering is how we're fighting back. But here's the thing - we've been fighting with one hand tied behind our backs. We've been encoding knowledge into runbooks that go stale, documentation that drifts, and tribal expertise that walks out the door when someone takes a new job. What if the platform itself could think alongside us? That's what we mean by agentic platform engine...
When Infrastructure Scales But Understanding Doesn’t
We all know this, even if we don't like to admit it: modern infrastructure can scale infinitely, but human understanding doesn't. We've all seen it happen - organizations going from managing dozens of servers to thousands of containers, from deploying weekly to deploying hundreds of times per day, from serving thousands of users to millions. The technology handled the scale beautifully. The humans? Not so much. This is the first industry issue that platform engineering should be addressing: how do we manage infrastructure complexity that has outgrown not just individual cognitive capacity, but our collectiv...
From 150 Unread to Zero Stress: Automating Inbox Triage with MCP and GitHub Copilot
Taming the Noisy Inbox: How I Used MCP to Automate Email and Teams Triage How the Model Context Protocol (MCP) turns your AI coding assistant into a workplace productivity engine — connecting Microsoft 365 data to your terminal workflow. The Problem We All Share If you work in a customer-facing role, you know the feeling. You open your laptop on Monday morning and you’re staring at 150+ unread emails, dozens of Teams threads, and the creeping anxiety that something important is buried in there — a customer escalation, an exec ask, a deadline you forgot about. You start scrolling. You context-switc...
Azure Skills Plugin – Let’s Get Started!
Part 2 of the Azure Skills Plugin series Previously: Announcing the Azure Skills Plugin This post is all about getting you up and running. I won't go deep on capabilities, architecture or anything like that here - that's coming in future posts in this series. The goal is to get the plugin installed, verified, and ready so you can start exploring right away. The Azure Skills Plugin works across many agent hosts including GitHub Copilot CLI, VS Code, and Claude Code. Each install takes under 60 seconds. This post covers all three, what gets installed, and how to verify it's working. 👉 Plugin repo:...
The Human Scale Problem in Platform Engineering
We keep doing this thing where we solve a problem, celebrate the victory, then realize we've created three new problems we didn't even know existed. Remember when manually configuring servers was the bottleneck? So we built containers. Great! Now we're orchestrating thousands of them. Remember when monolithic deployments were too slow? So we built microservices. Fantastic! Now we're drowning in distributed system complexity. We solved manual infrastructure provisioning with infrastructure as code. Perfect! Now we're coordinating dozens of Terraform modules across environments and wondering how we got here. ...
Announcing the Azure Skills Plugin
Part 1 of the Azure Skills Plugin series Coding agents like GitHub Copilot and Claude Code are great at code, but getting your app to production on Azure is not just about writing code. Really, it is about making the right calls. Which service fits this app? Which SKU fits this workload? Should this be App Service, Container Apps, Functions, AKS, or something else entirely? What needs to be validated before deploy? Which permissions, quotas, and guardrails matter? That is exactly why skills are taking off: they give agents practical knowledge on demand instead of forcing them to guess. Want to try t...
Platform Engineering for the Agentic AI era
For the last decade, platform engineering has relied on explicit API interaction layers: CLIs, SDKs, pipelines, wrappers, and UI workflows that translate human intent into machine‑safe API calls. AI agents are now short‑circuiting much of that stack. By combining natural language understanding, reasoning, and direct access to API specifications and control schemas, agents can convert human intent directly into validated platform actions, often without a bespoke interaction layer in between. Nowhere is this shift more visible than in Infrastructure as Code (IaC) and pipeline workflows, where agents are increasi...
Measuring actual AI Impact for Engineering with Apache DevLake
If you want to skip the explain and get started super quick with adoption + impact insights, use gh-devlake to deploy a GitHub Copilot impact dashboard in a few CLI commands. So! You've rolled out GitHub Copilot to your engineering teams. You've got the built-in dashboards. You know how many seats are assigned, what the acceptance rates look like, which editors your teams prefer. Maybe you've even pulled the Copilot Metrics API and built some charts. But here's the question your VP of Engineering or CTO is actually asking: "Is GitHub Copilot making us ship faster? Are we more reliable? Is code review g...
The OS for Intelligence: How GitHub Bridges the Fragmented AI Landscape
We are currently living through the "fragmentation phase" of the AI revolution. If you’re a developer, you know the drill: You have Claude Code open for reasoning. You have ChatGPT open for logic checks. Then you drop into your terminal to actually build the thing—manually copy-pasting context between three different windows. We call this the "fragmentation tax." It kills momentum, breaks your flow, and frankly, it’s a waste of cognitive energy. For engineering leaders, it’s even worse. It’s a governance nightmare and a silent killer of velocity. GitHub’s answer isn't just another tool; it’s an Op...