Skip to content

Cloud graphing for AWS, Azure and GCP

Map the cloud. Scan it. Merge everything.

cloudg collects your infrastructure into a graph, runs the security scanners you already trust over the same accounts and merges every finding against 28 compliance frameworks. You can also skip scanning entirely and just map what you have.

Get started Python API Sample report

pip install "cloudg[aws]"
An example map. Red marks the path in from the internet.

Three ways in

  • 01 · full pipeline

    Collect, scan, report


    Every provider in parallel, all four scanners and every report format, in one command.

    cloudg run -p all --regions all
    

    The pipeline

  • 02 · ingest only

    Merge output you already have


    No cloud credentials and no scanner binaries. Point it at the files and get the same reports.

    cloudg ingest --prowler ./out/
    

    Ingesting existing output

  • 03 · map only

    Inventory, every service linked


    Typed edges such as INVOKES, ASSUMES_ROLE and PROTECTS, with no scanner involved.

    cloudg map -p aws --org
    

    Inventory mapping

The same run from Python

The CLI is a thin layer over CloudGEngine, so anything a command does you can do from Python, one phase at a time if you like.

cloudg ingest --prowler ./prowler-output/ --trivy ./trivy.json -o ./reports
ingest.py
from cloudg import CloudGConfig, CloudGEngine

engine = CloudGEngine(CloudGConfig())
result = engine.run_from_reports_sync(
    {"prowler": ["./prowler-output/"], "trivy": ["./trivy.json"]},
    output_dir="./reports",
)
print(result.total_findings, result.severity_breakdown)
print(result.report_paths)  # findings.json and report.html
docker run --rm \
  -v "$PWD:/app/input:ro" \
  -v "$PWD/reports:/app/reports" \
  ghcr.io/morpheuslord/cloudg:latest \
  ingest --prowler /app/input/prowler-output/ --trivy /app/input/trivy.json

What happens in a run

Collectors read each provider concurrently and hand assets and edges to a NetworkX graph. Scanners run in parallel against the same accounts, and their output is normalised into one finding model, deduplicated and matched to compliance controls. The same data then leaves as an HTML report, JSON, GraphML, an RDF ontology, RAG chunks and, if you want it, an MCP server for your AI assistant.

flowchart LR
  P["AWS, Azure, GCP"] --> C[Collectors]
  C --> G[(Graph)]
  P --> S[Scanners]
  S --> N[Normaliser]
  G --> N
  N --> R["report.html, JSON, SVG"]
  G --> O["Ontology and RAG"]
  N --> M["MCP server"]
  O --> M

Where to go next

  • Guides


    Install, authenticate, map an account, run the scanners and read the report.

    Introduction

  • CLI reference


    Every command and every flag, generated from the click definitions so it matches --help.

    All commands

  • Python API


    CloudGEngine and the modules behind it, with signatures, parameters and runnable examples.

    Python API

  • Examples


    Complete scripts for mapping, ingesting, graph exports and error handling, each with its output.

    Examples

  • MCP server


    74 tools for Claude Desktop, Claude Code, Cursor, VS Code or any other MCP client.

    MCP server

  • Sample report


    The interactive report.html for a synthetic three-cloud estate, with no account needed.

    Open the sample report