Drop any infrastructure config and watch three AI agents parse it, map its topology, and issue a security compliance verdict — in real time.
Parses any infrastructure format
Shadow Architect orchestrates a Parser, a Topology Researcher, and an isolated Compliance Judge — each with its own context, tools, and responsibility. You get expert output at every layer, not a single overloaded prompt.
Drop a CloudFormation YAML, Terraform HCL, Docker Compose, or just describe your stack in plain English. The AI translates it into an interactive React Flow graph with auto-layout in seconds.
The Compliance Judge is a fresh LLM call with zero memory of the parsing phase. It receives only the serialized topology and issues a strict PASS, REVISE, or REJECT verdict with severity-tagged findings — no motivated reasoning.
Re-run the same config after a change and watch NEW, MODIFIED, and DELETED badges appear on the graph. The cost calculator applies live AWS pricing. The AI chat answers questions about your infrastructure using the full topology as context.
DRIFT DETECTION — Run #2 vs Run #1
Agent Architecture
Reads raw config — CloudFormation, Terraform, or plain English — and extracts a structured Resource list with types, names, and properties.
Determines how resources connect: security groups, subnets, data flows, IAM grants. Outputs a dagre-laid-out node graph.
Receives only the final topology — zero memory of earlier agents. Issues PASS / REVISE / REJECT with severity-tagged findings.
Day 2 Operations
Shadow Architect is designed for iterative infrastructure work — not just one-shot audits.
Re-run any config after changes. NEW, MODIFIED, and DELETED badges appear on graph nodes and pulse with color to signal what changed between runs.
Static AWS pricing is applied to every detected resource. See estimated monthly cost per node and total stack cost, with expandable line items.
Chat with the parsed topology using natural language. The AI knows every resource, connection, and finding — no copy-pasting JSON required.
No setup. Drop a config file or describe your stack in plain English.