TOOLS4AGENTSHow to

Azure Enterprise Infra Planner

Azure Enterprise Infra Planner: Set It Up With Claude Code

1,662
GitHub stars
4,353
Installs
2026-11-20
Last commit
Published 2026-11-21 · Last verified 2026-11-21

Learn how to set up the Azure Enterprise Infra Planner skill so your coding agent can plan and generate enterprise Azure infrastructure.

Planning enterprise Azure infrastructure means juggling networking, identity, security, compliance, and multi-resource topologies at once. A single missed pairing constraint or an insecure default can undo a whole design. The Azure Enterprise Infra Planner is a skill that gives your AI coding agent a structured way to work through that planning, then generate the Bicep or Terraform to match.

What it is

The Azure Enterprise Infra Planner is a skill published by Microsoft. Its description says it architects and provisions enterprise Azure infrastructure from workload descriptions. It is aimed at cloud architects and platform engineers who plan networking, identity, security, compliance, and multi-resource topologies with WAF alignment.

In plain words, it gives your agent a repeatable process. Instead of asking the agent to invent an Azure design from scratch, the skill walks it through a fixed pipeline: extract insights, research best practices, research resources, generate a plan, get your approval, generate infrastructure as code, then deploy. The output is Bicep or Terraform generated directly. The skill description says it does not use azd.

The skill also tells the agent which tools to reach for. It lists MCP tools for insights, best practices, the Well-Architected Framework service guide, Microsoft Learn search and fetch, and Bicep schemas. It lists CLI commands for deployment, Bicep builds, resource listing, and the Terraform workflow, plus checkov for security scanning.

The skill ships with reference files: a workflow document, a WAF checklist, a resources folder, a constraints folder, a schema for the plan output, and a referenced-workload document for brownfield cases. The main skill file points the agent to the workflow document as the place to start.

Installs compared with similar skills
Installs compared with similar skills, in installs
Nameinstalls
Azure Enterprise Infra Planner
4,353
Azure Kubernetes
9,895
Unit: installs. Source: skills.sh install counts, from each entry's stats.

Why it matters

The sources make several claims about why this structure helps.

First, the pipeline is gated. Every phase advances only after its gate passes. That means the agent does not jump from a vague description straight to deployed infrastructure. It has to clear each stage first.

Second, Phase 5 requires explicit user approval. You see the plan before any code is generated. If you do not approve, the pipeline loops back to plan generation.

Third, Phase 6 is described as a hardened, self-verifying gate. The generated infrastructure as code must be secure by default, pass local validation with az bicep build or terraform validate with zero errors, and pass a checkov security scan with no unresolved high or critical findings. The skill must also show the command output and emit a completion self-check before moving on.

Fourth, Phase 7 requires an explicit, risk-acknowledged deploy confirmation. Deployment does not happen silently.

Fifth, the skill handles existing environments. When you supply something that already exists, the skill inventories and references those resources and never recreates them. Phase 7 then deploys additively, incremental only, never modifying or destroying the referenced resources.

Finally, the skill is explicit about when to use it and when not to. Its description says to prefer azure-prepare for app-centric workflows. That keeps the tool focused on enterprise infrastructure planning rather than general app deployment.

Step by step

The skill file does not give a separate install command. What it does give is the setup your agent needs: the skill itself, the MCP tools it calls, and the CLI commands it runs. Here is how to put those pieces in place.

  1. Add the skill to your project so your coding agent can load it.The skill is defined by its SKILL.md file with the name azure-enterprise-infra-planner. Place it where your agent reads skills from.
  1. Confirm your agent can see the skill by its name and description.The description lists the trigger phrases, such as planning Azure infrastructure, architecting a landing zone, designing a hub-spoke network, planning multi-region DR topology, setting up VNets, firewalls and private endpoints, subscription-scope Bicep deployment, and Azure Backup for VM workloads.
  1. Make the MCP tools available to your agent.The skill lists six: insights_get, get_azure_bestpractices_get, wellarchitectedframework_serviceguide_get, microsoft_docs_fetch, microsoft_docs_search, and bicepschema_get. Each one serves a specific purpose in the pipeline.
  1. Check that the Azure CLI is available for Bicep work.The skill uses these commands during validation and deployment:
az bicep build
az deployment group create
az resource list
  1. If you plan to generate Terraform instead of Bicep, check that Terraform is available.The skill uses these commands:
terraform init
terraform validate
terraform plan
terraform apply
  1. Make checkov available.Phase 6 runs a checkov security scan and requires no unresolved high or critical findings before the pipeline advances.
checkov
  1. Point your agent at the workflow document.The skill says to follow the step-by-step instructions in workflow.md to execute the 7 phases. That file is the starting point for the actual run.
  1. Know where the skill writes its files.Artifacts go under your project root: .azure/insights.json in Phase 1, .azure/infrastructure-plan.json in Phase 4 with status moving from draft to approved to deployed, and infra/main.bicep plus infra/modules/* or infra/main.tf plus infra/modules/** in Phase 6.

A worked example

Suppose you are a platform engineer who needs a hub-spoke network with firewalls and private endpoints, plus a disaster recovery topology across regions. You open your coding agent and describe the workload.

The agent activates the skill because your request matches its trigger phrases. It first triages your input. If you have not referenced anything that already exists, it treats the work as greenfield and runs the phases straight through. If you point at a live resource group or an existing infrastructure plan, it treats the work as referenced and first inventories those resources, assigns roles, and marks them to be referenced rather than recreated.

In Phase 1, the agent extracts insights. It can call insights_get to retrieve insights about your existing Azure environment to guide planning decisions. It writes the result to .azure/insights.json.

In Phase 2, it researches best practices. It can call get_azure_bestpractices_get for Azure best practices on code generation, operations, and deployment, and wellarchitectedframework_serviceguide_get for the WAF service guide for a specific Azure service.

In Phase 3, it researches resources. It can call microsoft_docs_search to search Microsoft Learn for relevant documentation chunks, microsoft_docs_fetch to pull the full content of a Learn page by URL, and bicepschema_get for the Bicep schema definition of any Azure resource type at the latest API version.

In Phase 4, it generates the plan and writes .azure/infrastructure-plan.json with status draft. The plan follows the output schema in schema.md.

Phase 5 is your gate. The agent shows you the plan and asks for approval. If you say no, it goes back to Phase 4. If you approve, the plan status moves to approved.

In Phase 6, the agent generates the infrastructure as code. It writes either infra/main.bicep with modules, or infra/main.tf with modules. Then it validates. For Bicep it runs az bicep build. For Terraform it runs terraform validate. If validation returns errors, it fixes the code and re-validates. It also runs checkov and must clear any high or critical findings. It shows you the command output and emits a completion self-check.

Phase 7 is the deploy gate. The agent asks for an explicit, risk-acknowledged confirmation. For a greenfield deployment it runs az deployment group create or terraform apply. For a referenced deployment it deploys additively, incremental only, with a what-if preview, and does not destroy the referenced resources.

If something goes wrong along the way, the skill has an error table. A missing plan approval means the agent stops and prompts you. A pairing constraint violation, such as an incompatible SKU or resource combination, gets fixed in the plan before code generation. Missing files mean the agent checks that they exist at .azure/ and infra/ and re-creates them by following the workflow.

How it compares

Azure Prepare prepares azd-based Azure projects for deployment. It generates azure.yaml, infrastructure in Bicep or Terraform, and Dockerfiles for the Azure Developer CLI. The Infra Planner's own description says to prefer azure-prepare for app-centric workflows. So if your goal is getting an app deployed with azd, Azure Prepare is the better fit. The Infra Planner is for enterprise infrastructure planning and generates Bicep or Terraform directly, without azd.

Azure Deploy executes Azure deployments for already-prepared applications that have an existing .azure/deployment-plan.md and infrastructure files. It picks up where preparation ends. If your app is already prepared and you just need the deployment run, Azure Deploy is the better fit. The Infra Planner covers the planning and code generation that comes before deployment.

Azure Kubernetes plans, creates, and configures production-ready Azure Kubernetes Service clusters. If your workload centers on AKS, that skill is the better fit. The Infra Planner works at the broader infrastructure level, covering networking, identity, security, compliance, and multi-resource topologies.

Pitfalls

The skill depends on MCP tools. If a tool call times out, hits a connection error, or the tool does not exist, the skill retries once, then falls back to reference files and notifies you. You may lose some of the live research in that case.

Phase 5 will not let you skip approval. If the plan status is not approved, the agent stops and prompts you. Plan on reviewing the plan.

Phase 6 will not let you skip validation. The generated code must pass az bicep build or terraform validate with zero errors and a checkov scan with no unresolved high or critical findings. If those do not clear, the pipeline does not advance.

Phase 7 requires an explicit, risk-acknowledged deploy confirmation. Deployment is a deliberate step, not an automatic one.

Pairing constraints can block progress. If the plan contains an incompatible SKU or resource combination, it must be fixed in the plan before code generation.

Files must land in the right place. If the plan or code files are missing from .azure/ or infra/, the skill says to re-create them by following the workflow exactly.

The skill is scoped to enterprise infrastructure. Its description says to prefer azure-prepare for app-centric workflows, so it is not the right tool for every Azure task.

FAQ

What does this skill actually produce?

It produces a plan at .azure/infrastructure-plan.json and infrastructure as code at infra/main.bicep with modules, or infra/main.tf with modules. It can then deploy that code after your confirmation.

Does it need an API key?

The sources do not mention API keys. They do list the MCP tools the skill calls and the CLI commands it runs, so your setup needs those tools and commands available.

Can it touch my existing Azure resources?

Yes, but carefully. When you reference existing resources, the skill inventories and references them, never recreates them, and Phase 7 deploys additively, incremental only, without modifying or destroying them.

What happens if validation fails?

The agent fixes the generated code and re-validates. If the problem is unresolved, it notifies you. The pipeline does not advance past Phase 6 until validation is clean.

When should I use azure-prepare instead?

The skill's own description says to prefer azure-prepare for app-centric workflows. Use the Infra Planner when you are planning enterprise infrastructure such as landing zones, hub-spoke networks, or multi-region topologies.

Verdict

Use the Azure Enterprise Infra Planner if you are a cloud architect or platform engineer planning enterprise Azure infrastructure: landing zones, hub-spoke networks, multi-region disaster recovery, or subscription-scope deployments. It suits you if you want a gated process with plan approval, validation, a security scan, and a deliberate deploy step, and if you are comfortable having the MCP tools and CLI commands it relies on in place.

Skip it if your work is app-centric and azd-based. The skill's own description points you to azure-prepare for that. Skip it too if you want a one-shot deployment of an already-prepared app, where Azure Deploy fits better, or if your workload is centered on AKS, where Azure Kubernetes is the closer match.

Related on this site