> For the complete documentation index, see [llms.txt](https://help.cubelet.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://help.cubelet.ai/platform-and-integration/platform-and-integration.md).

# Platform & Integration

*Reference for developers and AI agents integrating with cubelet.ai's MCP surface.*

## What is cubelet.ai

cubelet.ai is an AI-native training and credentialing platform. *"Know the gap. Prove the judgment."* — the platform diagnoses what a learner doesn't yet know, closes that gap through mastery-paced study, and issues a verifiable credential once judgment against real scenarios is demonstrated.

cubelet.ai serves two audiences: learners, who study and practice through the product itself, and developers and AI agents, who integrate with the platform's content and tooling through MCP. **This section — Platform & Integration — is written for the second audience.** If you're looking for how to use the product as a learner, that content lives elsewhere on this site.

## The learner journey: Scaffold → Pursuit → Passport

Every learner moves through three stages:

* **Scaffold** — a gap diagnostic that assesses a learner's current mastery and surfaces their Gap: the delta between what they know and what a competency requires.
* **Pursuit** — mastery-paced training that closes the Gap, using Cubelets in the Library to study and Cases in the Lab to practice applied judgment.
* **Passport** — credential issuance. Once a learner has proven judgment (not just study activity), the Passport issues a W3C Verifiable Credential.

## Core taxonomy at a glance

These terms are locked and used consistently across the platform and its APIs. Full definitions live on the Glossary page — this is the quick-reference version.

| Term         | One-line definition                                                                                                                              |
| ------------ | ------------------------------------------------------------------------------------------------------------------------------------------------ |
| **Cubelet**  | A knowledge atom with 6 faces (WHAT, WHY, HOW, WHERE, WHEN, APPLY), mapped 1:1 to a competency.                                                  |
| **Case**     | A judgment exercise: a fictional organization, a set of practices, and ground truth, produced by the simulator pipeline. Never called a cubelet. |
| **Library**  | Where learners study Cubelets.                                                                                                                   |
| **Lab**      | Where learners practice with Cases.                                                                                                              |
| **Catalog**  | The content registry. An item is registered only when both its Neo4j graph edges and its R2 content are written — a partial write doesn't count. |
| **Passport** | The credential surface; issues W3C Verifiable Credentials for proven outcomes.                                                                   |
| **Gap**      | The delta between a learner's current mastery and what a competency requires, surfaced by Scaffold.                                              |

See the Glossary for canonical, self-contained definitions of every term, including related concepts like Competency, Face, and Framework element ID.

## Domains available on the platform

cubelet.ai currently covers:

* **CMMC**
* **ISACA Tri-Cert** (CISA, CISM, CRISC, CDPSE)
* **AI Governance** (ISO 42001, EU AI Act, NIST AI RMF, TRAIGA)
* **NIST Privacy Framework**
* **VulnOps**
* **GRC Engineering**

Each domain has its own Cubelets (Library content) and, where a simulator has been built for it, Cases (Lab content).

## Federated MCP architecture

cubelet.ai doesn't expose one monolithic API. Instead, each capability area is served by its own MCP server behind its own endpoint — content and tooling for a given domain or function live on the server that owns it, rather than all being bolted onto a single surface. A client or agent connects to the specific server(s) it needs for the integration at hand.

See the MCP servers page in this section for the current server list, what each one is scoped to, and how to authenticate against it.

## What's in this section

* **Architecture** — how the federated MCP servers, the Catalog's dual-write model (Neo4j + R2), and the platform's data model fit together.
* **MCP servers** — the current server inventory: what each server is scoped to, its endpoint, and the tools it exposes.
* **Connecting a client** — how to point an MCP-capable client or agent at cubelet.ai's servers.
* **Reference** — supporting technical reference for building against the platform.