Diagrams That Prove Your Terraform Plan Isn't Fiction

What it does

Reads your Terraform code—or a simple JSON graph of resource names—and renders professional architecture diagrams with official AWS, Azure, and GCP icons. It resolves count, for_each, and modules by deriving the graph from terraform plan, so the picture reflects the real infrastructure. Output formats include PNG, SVG, PDF, interactive HTML, and editable draw.io.

The interesting bit

The project doubles as an MCP server and agent skill, letting Claude, Codex, Gemini, or Cursor generate diagrams directly from a conversation. A simplified view strips out VPCs and subnets for executive decks, while the interactive HTML output packs pan, zoom, and animated data flow into a single offline-capable file.

Key highlights

• Runs entirely client-side; no cloud credentials required and source code never leaves the machine.
• Supports 385 AWS, 264 GCP, and 245 Azure resource types with official icon sets.
• Accepts a terse JSON graph format using raw Terraform resource names, eliminating the need to learn a custom DSL.
• Optional AI annotations via local Ollama, AWS Bedrock, or any OpenAI-compatible endpoint—only metadata is sent, never .tf files.
• Terragrunt-aware and CI/CD friendly, with a pre-generated plan mode that decouples diagramming from live infrastructure access.

Caveats

• The docs inconsistently describe Azure coverage as both “core services” and “full support.”
• Terraform-native diagramming depends on a local Terraform 1.x or OpenTofu installation; JSON graph mode sidesteps this entirely.
• Optional AI annotations require their own backend setup—local Ollama, AWS credentials for Bedrock, or environment variables for a REST API.

Verdict

Cloud engineers tired of maintaining stale draw.io files will find this genuinely useful; teams already happy with Mermaid or who don’t diagram at all can skip it.