CALM

CALM — Common Architecture Language Model — is a FINOS specification for describing software architectures in a structured, machine-readable format. Where traditional architecture documentation lives in slide decks and diagram tools, CALM describes systems as typed, validated schemas: services, their interfaces, the flows between them, the security properties they should have, and the constraints they must satisfy. An architecture described in CALM is not a picture of a system — it is a definition that tooling can reason about, validate, and compare against policy.

CALM was originated at Morgan Stanley and contributed to FINOS, where it is maintained by an open community of contributors. The specification covers the full lifecycle of an architecture: initial design, validation against constraints, publication as a governed artifact, and ongoing drift detection as the real system evolves away from its approved description.

Architecture governance has historically been one of the weakest links in enterprise compliance programmes. A system gets designed, reviewed, and approved — and then the real implementation diverges from the approved design, slowly and incrementally, until the production system no longer resembles what the architecture review board signed off on. No one set out to circumvent the process; the process simply had no way of detecting drift.

CALM makes architecture a first-class compliance artifact. Because a CALM architecture is machine-readable, it can be compared automatically against a running system. Deviations are detectable. And because CALM architectures are validated against policy constraints at design time — before any code is written — compliance issues are caught when they are cheapest to fix, not when they are discovered in a production audit.

For regulated industries, this matters beyond internal governance. Frameworks like DORA and NIST require organisations to maintain accurate, current descriptions of their critical systems. CALM provides an architecture format that satisfies those requirements in a form that is both human-reviewable and machine-verifiable.

Meridian is a contributor to CALM under FINOS governance. Our work on the specification focuses on the intersection between architecture definitions and compliance policy — specifically, the mechanisms by which a CALM architecture can declare the policy constraints it must satisfy, and the tooling by which those constraints can be evaluated automatically.

We chose to build Meridian Loft on CALM rather than a proprietary architecture format because the value of architecture governance depends entirely on adoption. A proprietary format requires every architect in the organisation to learn a Meridian-specific tool. CALM, as an open standard with an existing community, is something architects can adopt independently of Meridian — and the governance and validation capabilities Meridian adds are layered on top of a specification they already know.

Meridian Loft is built entirely on CALM. When a Cloud Architect uses Meridian Loft to design a system, they are authoring a CALM architecture — one that Meridian Loft automatically validates against the policies active in Meridian Chancery before the design is approved. If a proposed architecture violates a security constraint or fails to satisfy a regulatory requirement, Meridian Loft flags it at design time. If it passes, the approved CALM definition is recorded as an immutable, auditable artifact.

This means that every system deployed through Meridian Slipway has a corresponding approved architecture on record — a design-time commitment that Meridian Patrol can check against at runtime. Drift between the approved CALM definition and the actual deployed state is detectable automatically, closing the loop between architecture governance and continuous compliance monitoring.