AZ-401 is the first Microsoft DevOps exam written for a world where an agent can open the pull request, run the pipeline, and file the incident. That does not make it a prompt-writing test. It is still a DevOps exam, with the same pipeline, identity, and observability judgments, plus a new layer of questions about how AI assistance fits inside those controls. Candidates who study only the new AI features miss the 35 to 40 percent of the exam that still lives in build and release pipelines.
TL;DR
- Exam: Designing and Implementing Microsoft Agentic DevOps Solutions (beta), code AZ-401.
- Credential: Microsoft Certified: Agentic DevOps Engineer Expert (beta). AZ-401 is the only required exam.
- Passing score: 700. Microsoft does not publish a question count, duration, or price for the beta.
- Beta behavior: Your exam is not scored immediately while Microsoft collects question-quality data.
- Largest domain: Build and release pipelines at 35 to 40 percent. Do not let the AI topics pull you away from it.
- New testable skills: MCP servers for Azure DevOps and GitHub, GitHub Agentic Workflows, Azure Copilot Observability Agent, and Azure SRE Agent.
- Preparation: Reading the study guide is not enough. Practice YAML, Bicep, KQL, Azure DevOps, GitHub Actions, and secretless authentication hands-on.
What the AZ-401 Actually Is
AZ-401 validates the design and implementation of AI-assisted DevOps solutions across Azure DevOps and GitHub. Microsoft's audience profile describes a developer or infrastructure administrator who understands people, processes, and products well enough to enable continuous delivery, and who now enhances that delivery with AI assistance tools.
The role is not a narrow automation specialist. Microsoft expects you to collaborate with developers, site reliability engineers, platform engineers, Azure administrators, and security engineers. The exam therefore mixes process design, source control, pipelines, security, and observability rather than focusing on a single tool.
The beta label matters for planning, not for preparation. Beta exams use the same blueprint and the same 700 pass mark, but Microsoft withholds scores while it studies how the questions perform. Question count, time limit, and price are not published on the exam page. If you need a predictable test-day timeline, treat AZ-401 as an early-adopter exam and plan for the beta process rather than a standard 100-minute slot.
Exam at a Glance
| Item | Detail |
|---|---|
| Exam code | AZ-401 |
| Exam title | Designing and Implementing Microsoft Agentic DevOps Solutions (beta) |
| Credential | Microsoft Certified: Agentic DevOps Engineer Expert (beta) |
| Level | Expert |
| Passing score | 700 |
| Duration | Not published for the beta |
| Question count | Not published for the beta |
| Price | Not published on the exam page |
| Languages | English |
| Prerequisite exams | None listed, AZ-401 is the only required exam |
| Practice assessment | Not currently available |
Microsoft says a practice assessment usually arrives within about eight weeks of an exam leaving beta and reaching general availability. The study guide and the exam sandbox are available now. The sandbox is worth ten minutes even before you start studying, because it shows the question formats you will meet on test day.
Microsoft also lists no learning paths and no instructor-led courses for this certification yet. That is normal for a new expert exam, but it changes your study plan. You cannot wait for a curated course. You have to build the plan from the study guide and the product documentation.
The Five Domains and Their Weights
The blueprint is deliberately uneven. One domain carries more than a third of the exam, and two others are only 10 to 15 percent each.
| Domain | Weight |
|---|---|
| Design and implement build and release pipelines | 35 to 40% |
| Design and implement processes and communications | 15 to 20% |
| Design and implement a security and compliance plan | 15 to 20% |
| Implement an instrumentation strategy | 15 to 20% |
| Design and implement a source control strategy | 10 to 15% |
Build and release pipelines is the center of gravity. Microsoft lists package management with Azure Artifacts and GitHub Packages, semantic and calendar versioning, quality and release gates, testing strategy, Azure Pipelines and GitHub Actions, GitHub Agentic Workflows, deployment strategies, infrastructure as code, and pipeline maintenance. If you are short on time, this is the domain that decides your score.
Processes and communications covers traceability and flow of work in Azure Boards, success metrics, documentation with AI-assisted tools, agentic development requirements such as agents, skills, and loops, MCP servers for Azure DevOps and GitHub, and Azure Boards plus GitHub integration. This is where the exam tests whether you can design a delivery process that an AI agent can operate inside, not just a process a human can follow.
Source control is the smallest domain, but it is not optional. It covers branching strategies including trunk-based, feature, release, Git Worktree, Git flow, and GitHub flow; pull request workflows; branch policies and protection rules; repository permissions and tags; Git hooks; and migration from Azure DevOps to GitHub. The questions tend to be scenario-based, so memorizing the name of a strategy is less useful than knowing when it breaks down.
Security and compliance covers service principals and managed identities, GitHub Apps and GITHUB_TOKEN, secretless authentication, Azure Key Vault, secrets management, GitHub Advanced Security, and Microsoft Defender for Cloud DevOps Security. The theme is removing long-lived credentials from the delivery path without breaking the pipeline.
Instrumentation covers dashboards and DevOps metrics, Azure Monitor and Log Analytics integration, alerts, Azure Copilot Observability Agent, Azure SRE Agent, Application Insights distributed tracing, GitHub Copilot agent logs, KQL, and natural language log queries. This domain rewards candidates who have actually queried a workspace rather than only read about one.
What Is New Compared With Older Azure DevOps Exams
The familiar AZ-400 material still appears: YAML pipelines, environments, approvals, branch policies, Key Vault, and Application Insights. The difference is the agentic layer.
Microsoft now names MCP servers for Azure DevOps and GitHub as testable material. MCP, the Model Context Protocol, is how an AI client discovers and calls tools. In an exam scenario, the question is usually about giving an agent a controlled interface to a system rather than handing it a personal access token. That connects directly to the security domain.
GitHub Agentic Workflows is another new topic. The exam expects you to understand how agent-driven automation fits into a repository workflow, including where human approval and policy checks still belong. The same pattern appears in the processes domain, where Microsoft lists agents, skills, and loops as requirements to design for.
Observability has changed too. Azure Copilot Observability Agent, Azure SRE Agent, and GitHub Copilot agent logs are now in scope. The exam is not asking you to configure every feature. It is asking whether you can design an instrumentation strategy that still gives you evidence when an agent, not a person, performed the action.
The security implications are the part candidates underestimate. An agent with a broad token is a new class of insider risk. Microsoft's answer in the blueprint is secretless authentication, managed identities, GitHub Apps, GITHUB_TOKEN scoping, Key Vault, GitHub Advanced Security, and Defender for Cloud DevOps Security. Expect questions that ask you to choose the least-privilege option, not the fastest one.
Who Should Take AZ-401
Microsoft's audience profile is explicit: you should have strong skills in either administering or developing solutions on Microsoft technologies, practical experience with YAML, KQL, and Bicep, and experience implementing both Azure DevOps and GitHub solutions using Microsoft and GitHub AI tools.
That last clause is the one to take seriously. If you have only used Azure Pipelines, spend time in GitHub Actions and GitHub Advanced Security. If you have only used GitHub, spend time in Azure Boards, Azure Artifacts, and Azure Pipelines environments. The exam deliberately spans both platforms because the job does.
The credential is aimed at DevOps engineers, developers, platform engineers, and solution architects. It is an expert-level exam, so Microsoft assumes you already know the mechanics. The questions are about design choices and trade-offs, not definitions.
There is no listed prerequisite certification. That does not mean the exam is entry-level. It means Microsoft is not forcing you through AZ-104 or AZ-400 first. If you are new to Azure administration, build that foundation before attempting AZ-401.
A Study Plan That Matches the Blueprint
Start with the official AZ-401 study guide and copy its five domains into a checklist. Then allocate study time in proportion to the weights, not in proportion to how interesting the new AI topics are.
A practical split looks like this:
- Pipelines first. Build a YAML pipeline in Azure DevOps and an equivalent GitHub Actions workflow. Add environments, approvals, and a deployment strategy. Then break it and fix it.
- Security second. Replace a stored secret with a managed identity or a federated credential. Scope a GITHUB_TOKEN. Configure a Key Vault reference. Read the GitHub Advanced Security and Defender for Cloud DevOps Security documentation.
- Process and source control together. Map a work item to a branch, a pull request, a pipeline run, and a deployment. Then design the same flow for an agent that opens the pull request itself.
- Instrumentation last, but do not skip it. Query a Log Analytics workspace with KQL. Build one alert. Trace one request through Application Insights. Then read what the Copilot Observability Agent and SRE Agent actually do.
- Practice under time pressure. The exam is scenario-based, so recognition is not enough. You need to make a design decision and move on.
Community pass reports for this exam are still thin because it is in beta, so treat any specific question count or difficulty claim online with skepticism. The study guide and the product documentation are the reliable sources.
A Worked Scenario
Consider a question in the style Microsoft uses:
A team wants an AI agent to triage failed production deployments and open a pull request with a proposed fix. The agent needs to read pipeline logs, inspect the repository, and create a branch. Which design should you recommend?
The tempting answer is to give the agent a personal access token with repository and pipeline access. That works, and it is the wrong choice for an expert exam.
The better answer separates identity from permission. Use a GitHub App or a managed identity with narrowly scoped permissions, store no long-lived secret in the pipeline, and let the agent read logs through a controlled interface such as an MCP server. Require a human approval before the pull request merges, and log the agent's actions so the instrumentation strategy can reconstruct what happened.
That single scenario touches the security, process, source control, and instrumentation domains at once. Questions in this exam are built the same way. The answer is rarely a single product. It is the design that keeps the delivery path observable and least-privileged while still letting the agent do useful work.
Practice and Next Steps
CertCompanion's AZ-401 practice exam has 400 questions weighted to the five domains, with 30 free to start. Use it to find the domain that is actually weak rather than the one you enjoy studying. The questions are written to make you choose between plausible designs, which is the skill the exam rewards.
If you are still deciding between Microsoft's DevOps credentials, the Microsoft certification hub lists the current exams and how they connect. AZ-401 is the expert-level agentic path. AZ-400 remains the established DevOps Engineer Expert exam. Check which one your employer or target role recognizes before you book.
For AZ-401 itself, the sequence is simple: confirm you can build and secure a pipeline on both platforms, confirm you can instrument it, then practice the design decisions until they are automatic. The beta window is an opportunity to be early, not a reason to prepare less.
Try 3 AZ-401 practice questions
Pick an answer to see if you are right and read the explanation.
1. A pipeline fails because a secret is not masked in logs. The team is using a standard variable in the pipeline definition. Which variable type should they use to ensure the secret is masked? (Select one!)
Explanation
Secret variables in Azure DevOps are automatically masked in logs to prevent leakage. Standard variables are not masked. Environment variables and runtime variables do not inherently provide masking capabilities unless explicitly configured as secret variables. Using secret variables is the standard practice for handling sensitive data in pipelines.
2. A development team is publishing internal .NET libraries to Azure Artifacts. They require a versioning strategy that uses the year and month for the major and minor versions. Which versioning scheme should they adopt? (Select one!)
Explanation
CalVer (Calendar Versioning) uses the year and month for the major and minor versions, aligning with the requirement. SemVer uses major, minor, and patch numbers based on compatibility, not dates. 'byPrereleaseNumber' is a specific Azure Pipelines task setting for date-based prerelease labels, not a general versioning strategy for the library itself. 'byBuildNumber' uses the pipeline run ID, not calendar dates.
3. A platform engineering team is configuring the Azure Copilot Observability Agent to analyze Azure Firewall application rule decisions. Which Log Analytics table should they query to retrieve this data? (Select one!)
Explanation
The 'AZFWApplicationRule' table stores Azure Firewall application rule decisions, which include HTTP/HTTPS traffic filtering. 'AZFWNetworkRule' stores network rule decisions (IP/port filtering), 'AZFWThreatIntel' stores threat intelligence matches, and 'AZFWNatRule' stores NAT rule decisions. To analyze application rule decisions, the team must query the 'AZFWApplicationRule' table.