Modern infrastructure teams face a paradox.

We have powerful tools like Terraform, cloud-native services across AWS, Azure, GCP, and IBM Cloud, and managed platforms such as Terraform Cloud—yet infrastructure delivery remains slow, fragmented, and heavily dependent on specialist knowledge.

At the same time, organizations are under pressure to:

• Move faster

• Reduce cloud spend

• Maintain compliance

• Support multi-cloud strategies

To address this, we built an agentic AI-based Intelligent Infrastructure Orchestrator, powered by WatsonX Orchestrate, Terraform Cloud, and a set of specialized infrastructure agents. The result is a system that transforms natural language intent into secure, governed, and production-ready infrastructure—across clouds.

System Architecture

At a high level, the system is composed of:

1. Supervisor Agent

• Intent classification and routing

• Context preservation across turns

• Safety checks and execution gating

• Coordination across agents

2. Specialized Agents

• Requirement Analysis Agent – architecture inference

• Knowledge Base Agent – Terraform Registry intelligence

• Terraform Cloud Management Agent – workspace & run automation

• No-Code Deployment Agent – minimal-input infra provisioning

• Cost & Utilization Agent – post-deployment visibility

3. Terraform MCP Server

A critical abstraction layer that:

• Translates agent decisions into executable Terraform actions

• Interfaces with Terraform Cloud APIs

• Pulls metadata from Terraform Registry

• Connects to cloud provider tooling

This ensures LLMs never talk directly to cloud APIs.

Below is a brief visual representation of this system:

Terraform Cloud as the Execution Control Plane

Terraform Cloud is not just a backend—it’s the enforcement layer connected via the official Terraform MCP Server deployed remotely.

The Terraform Cloud Management Agents automates:

• Org / project discovery

• Workspace lifecycle

• Variable and secret injection

• Run orchestration

Key benefits:

• All infra changes and checks can be performed in natural language using watsonx orchestrate

• Seamless multi-cloud executions

Terraform Cloud becomes the single source of truth, while agents act as intelligent operators.

Terraform Registry as a Queryable Knowledge Graph

Terraform Registry is treated as a live knowledge source, not static documentation.

The Knowledge Base Agent:

• Queries modules and providers

• Evaluates compatibility and maturity

• Surfaces required inputs/outputs

• Provides secure configuration patterns

This significantly reduces time spent searching, validating, and copy-pasting examples. 

Post-Provisioning: Cost and Utilization Intelligence

Infrastructure orchestration doesn’t end at deployment.

The Cost & Utilization Agent:

• Pulls active resource inventory from AWS and IBM Cloud

• Correlates usage with billing data

• Flags underutilized or idle resources

• Produces structured summaries and visual outputs

This closes the feedback loop:

• Provision → Observe → Optimize

From a systems perspective, this turns the platform into a continuous infrastructure optimization engine, not just a provisioning tool.

Demo Walkthrough