PROJECT 001 / Agentic AI Engineering
● In Development

Azure AI Infrastructure Engineer.

Building an intelligent multi-agent system for Azure infrastructure design, Infrastructure as Code automation, security assessment, and human-approved deployment workflows.

AzureClaudeAgentic AITerraformPythonMCP
01 / PROJECT OVERVIEW

From requirements to infrastructure.

Traditional cloud infrastructure engineering involves architecture design, configuration generation, security reviews, deployment planning and validation.

This prototype explores how multiple specialised AI agents can collaborate across these tasks while preserving deterministic validation, least-privilege access and human control over infrastructure changes.

TARGET CAPABILITIES
  • Discover scoped Azure infrastructure
  • Translate requirements into architecture specifications
  • Generate Terraform configurations
  • Identify security and compliance risks
  • Prepare reviewed deployment plans
  • Validate approved infrastructure changes
02 / SYSTEM DESIGN

Multi-Agent Architecture

A planned orchestration model separating reasoning, execution tools, validation and deployment approval.

USER REQUESTNatural Language Infrastructure Requirements
COORDINATION LAYERClaude OrchestratorTask routing · Agent handoffs · Workflow state

Architect Agent

Designs Azure architecture and generates structured specifications.

IaC Engineer

Generates and validates Terraform configurations.

Security Reviewer

Reviews infrastructure security, policies and IaC findings.

DevOps Engineer

Prepares CI/CD workflows, plans and deployment artifacts.

Validation Engineer

Checks deployment state, configuration and expected outcomes.

CONTROLLED TOOL LAYERMCP · Azure SDK · Terraform · Security Scanners
External Human Approval GateRequired before infrastructure deployment
CONTROLLED EXECUTIONIsolated Deployment Runner + Post-Deployment Validation

Conceptual target architecture. Agents do not receive unrestricted deployment credentials. The approval system must be implemented outside the model.

03 / ENGINEERING STACK

Tools & Technologies

Core technologies planned for the prototype.

Claude Agent SDK
Model Context Protocol
Microsoft Azure
Python
Terraform
Azure Resource Graph
Checkov
GitHub Actions
04 / SAFETY BY DESIGN

Autonomy with boundaries.

AI agents assist with engineering decisions, but sensitive infrastructure operations require deterministic controls.

  • Read-only discovery as the starting point
  • Resource-group-scoped lab permissions
  • Static checks and Terraform plan validation
  • Human approval enforced outside the AI agent
  • Audit logging and explicit rollback procedures
  • No unrestricted autonomous production deployment
05 / LEARNING ROADMAP

Eight Labs. One Engineering System.

A progressive learning path from development setup to a controlled end-to-end prototype.

LAB 01Planned

Development Environment

Set up Python, Claude, Azure CLI and the project repository.

LAB 02Planned

Azure Discovery Agent

Connect read-only Azure inventory tools through MCP.

LAB 03Planned

Multi-Agent Orchestration

Coordinate specialist agents with structured handoffs.

LAB 04Planned

Terraform Engineering Agent

Generate and validate infrastructure as code.

LAB 05Planned

Security & Compliance Review

Implement static checks and security feedback loops.

LAB 06Planned

Controlled Deployment

Introduce verified plans and external approval gates.

LAB 07Planned

CI/CD & Observability

Integrate GitHub Actions, logging and audit trails.

LAB 08Planned

End-to-End Capstone

Evaluate the complete workflow and document limitations.

Full technical instructions and tested code examples will be published in the documentation.

Open Learning Path →