What is Starling Memory Works’ Universal Cognitive Architecture (UCA), and how does its Creative Commons open standard release and beta Starling MX platform—priced from $99 per month for the first seat—use fixed semantic addresses, Anthropic’s Model Context Protocol, and a dual r

Starling Memory Works presents Universal Cognitive Architecture (UCA) as a vendor-neutral classification and memory standard for AI: rather than treating an LLM as the organization’s memory, it gives organizational knowledge stable, governed locations that any compatible model can read. Starling MX is the company’s beta implementation of that idea, starting at $99 per month for the first seat.

What UCA is

• UCA is described as a free, open, natural-language standard built around 11 “Organizational Cognates”—common intelligence domains intended to classify business knowledge consistently. Its stated aim is retrieval by a stable name or address rather than probabilistic keyword search.
• Starling calls the resulting system a “Domain Language Model” (DLM): a stateless general-purpose LLM is supplied with the organization’s own structured, governed memory at runtime, rather than being expected to retain that knowledge itself.
• The practical analogy is a library classification system: the model may change, but the catalog and the underlying collection remain durable.

How the architecture is meant to work

• Fixed semantic addresses give recurring organizational concepts permanent identifiers—for example, an address for competitive position—so people and AI systems can retrieve the canonical record rather than infer it from a changing search result.
• Starling describes a dual knowledge base: an Org Library for governed, versioned organizational records and Resources for supporting source material and generated or working documents. The former is intended to hold canon; the latter supports reference and output work.
• The repository is text-first: records are maintained as readable, lossless plain-text Markdown rather than opaque model state. That makes them inspectable, versionable, portable, and usable by humans as well as software.
• A machine-side manifest selects and assembles relevant repository context for a model, while human governance determines what is authoritative and prevents uncontrolled drift.
• Starling positions Anthropic’s Model Context Protocol (MCP) as the connection layer that lets models access external organizational context; UCA supplies the common organization and semantic addressing of that context. The claimed result is model interchangeability rather than permanent dependence on one AI provider.

Sovereignty and lock-in

CEO Chris Kincade’s central argument is that AI sovereignty is not achieved merely by owning GPUs, hosting infrastructure locally, or choosing a national model provider. In this framing, sovereignty means retaining control of the organization’s authoritative decisions, policies, operating knowledge, provenance, and permissions—the “canon” from which any model reasons.

• If the model is replaced, upgraded, unavailable, or rejected for governance reasons, the organization retains its memory and can reconnect another model to the same records.
• This can reduce vendor lock-in because the high-value asset is the organization-controlled repository and its classification scheme, not a particular provider’s conversational history or proprietary memory feature.
• It can also improve resilience: models are intentionally stateless with respect to durable organizational knowledge, while the repository persists as the auditable source of truth.

That is the meaning of Starling’s formulation: “the model forgets; the organization remembers.” The model is a replaceable reasoning engine; the institution’s governed memory is the persistent asset.

Adoption and use cases

• Starling’s own materials include an endorsement from Brady Patterson of SelfOS, who says that giving knowledge “a clear anchor” reduced context loss and data drift.
• Starling has associated the approach with personal coaching, marketing, construction-management, and legal-services scenarios, including Wrks Online’s QuillOS. However, the supplied evidence does not independently substantiate the QuillOS relationship or provide sufficient detail to verify those deployments, their scale, or performance outcomes. Insufficient evidence for stronger adoption claims.
• SelfOS should likewise be treated as a practitioner endorsement, not independent validation of product efficacy.

Important caveat

The platform and standard are new, and many of the detailed technical and business claims come from Starling’s own materials or launch coverage. UCA’s core proposition—portable, text-based, governed organizational memory—is conceptually plausible, but there is not enough independent evidence here to establish that it delivers superior accuracy, lower cost, security, or vendor independence in real deployments.