Architecture-as-Code Reference Platform (open prototype)

Problem

Architecture-as-Code is easy to agree with and hard to picture. Leaders and engineers needed to see what an architecture control plane actually looks like before committing to one.

Approach

• Modeled systems, patterns, standards and waivers as machine-readable, version-controlled catalogs (YAML, JSON and C4/Structurizr).
• Built an operational lifecycle (plan, build, test, accept, deploy) with automated schema validation, compliance checks and generated diagrams and documentation, gated by pull requests.
• Shipped developer utilities: a CLI (aac init, validate), an MCP server for IDEs and AI assistants, an OPA/Rego policy engine, context-driven development with Copilot Spaces and a reference GitHub Actions pipeline.

Governance

• Executable guardrails with pass/fail results on every pull request, including encryption, autoscaling and API-gateway policies.
• Waivers as first-class records with risk, expiry and remediation tracking.
• Model-to-runtime drift detection and architecture decision records kept next to the code.

Outcome

• An executive scorecard computed directly from the artifacts: standards conformance, risk posture, technical-debt exposure, remediation velocity, pattern reuse and cloud spend at risk.
• A public reference that colleagues and the community can explore, referenced in my most-read Architecture-as-Code writing.