FRAMEWORKS
The Sparx Systems architecture platform is a layered stack of components: each adding capability on top of the last. Most organizations that have been running Sparx EA for any length of time already have the first three layers: the EA desktop client, the shared EA Repository database, and Pro Cloud Server for collaboration and web access. EA GraphLink is the fourth layer: the new capability that opens the repository to AI and BI tools. Kernaro is the fifth: a stakeholder intelligence platform that rides on top of EA GraphLink to deliver natural language access to the architecture. Understanding how these layers fit together is the prerequisite for making sound purchasing and implementation decisions.
This page explains each component, what it does, what it requires, and how they integrate: from the perspective of organizations that are evaluating the platform, expanding their current deployment, or trying to understand what they need to unlock specific capabilities.
| Component | Purpose | Requires | Who Uses It |
|---|---|---|---|
| Sparx EA (desktop client) | Modeling tool: create, edit, navigate the architecture model | Windows OS | Architects, modellers |
| EA Repository | Shared database: team collaboration, version control, model storage | SQL Server / MySQL / PostgreSQL | All EA users (via client) |
| Pro Cloud Server (PCS) | Middleware: browser access, security, cloud hosting, integrations | EA Repository | Architects (browser), stakeholders (read) |
| EA GraphLink | AI/BI connectivity: GraphQL API and MCP Server for the repository | EA Repository + PCS | BI tools, AI platforms, developers |
| Kernaro AI Hub | Stakeholder intelligence: natural language AI interface to the architecture | EA GraphLink | Business stakeholders, executives, non-architects |
| Kernaro Assist | Architect productivity: in-EA AI assistant for modeling tasks | Sparx EA (no GraphLink required) | Architects |
Sparx EA is a Windows desktop client produced by Sparx Systems (Melbourne, Australia). It is the primary authoring environment for enterprise architects: the tool where models are created, diagrams are drawn, and repository content is managed.
What it does:
Sparx EA supports over a dozen modeling languages and notations: UML 2.5.1, SysML 1.x, ArchiMate 3, BPMN 2.0, ERD, TOGAF ADM, and others: within a single repository. Every element created in Sparx EA is a model element: it has properties, relationships, tagged values, and can appear on multiple diagrams. This is the fundamental difference between Sparx EA and a diagramming tool like Visio: the model underlies the diagrams, not the other way around.
Key capabilities:
What it is not:
Sparx EA is not a browser-native tool. It is a Windows desktop application. It is not a stakeholder portal. It does not provide a consumer-friendly interface for non-architect users. These gaps are addressed by the layers above.
Editions:
Sparx Systems offers several editions: Architect (single user, limited repository access), Professional (full feature set), Corporate (shared repository), Ultimate (full feature set including simulation). For team deployments, Corporate or Ultimate is standard. Licensing is available as perpetual plus maintenance, or subscription.
Licensing note: Sparx EA software is purchased directly from Sparx Systems or through authorized resellers. Sparx Services does not sell the software. We provide the bill of materials, advise on edition and quantity, and implement and configure the platform. This applies to all Sparx Systems software components (EA, PCS, EA GraphLink).
The EA Repository is a relational database that stores all model content for a team deployment of Sparx EA. When architects work on a shared model, they are all reading from and writing to this database.
Supported databases:
What the repository stores:
Every model element, diagram, relationship, tagged value, document, and diagram image in Sparx EA is stored in the repository database. The repository schema is documented by Sparx Systems and is directly accessible: which is the foundation for EA GraphLink’s connectivity.
Repository management:
The repository requires standard database administration: backups, performance tuning, user access management, and version management (Sparx EA uses element locking for concurrent access). For cloud-hosted deployments, the database runs in a managed cloud database service (Azure SQL, AWS RDS, etc.).
Most organizations that have been running Sparx EA in team mode for more than a year already have an EA Repository configured. If the repository has been well-maintained and the model is well-structured, adding EA GraphLink on top is a relatively straightforward addition. If the repository has accumulated diagram debt or inconsistent modeling, MDG governance remediation is typically the first step.
Pro Cloud Server (PCS) is middleware produced by Sparx Systems that sits between the EA Repository and a range of client interfaces. It is a server-side application that runs on Windows Server.
What PCS does:
Repository gateway: PCS acts as the primary connection point between EA clients and the repository database: abstracting the database connection string from clients, managing connection pooling, and enabling centralized access control.
Browser-based access: PCS includes a built-in web server that delivers Sparx EA’s browser-based interface (WebEA and Prolaborate integration points). This allows stakeholders to browse the architecture model from any browser without installing the desktop client. Browser access provides read capability and limited editing: full authoring requires the desktop client.
Security and authentication: PCS handles authentication (including integration with Active Directory/LDAP and SAML-based SSO), authorisation (which users can access which repositories), and audit logging. This is the enterprise security layer for Sparx EA deployments.
Cloud hosting enablement: For cloud-hosted EA deployments (Azure, AWS, or private cloud), PCS is the component that makes the repository accessible from anywhere: enabling remote architect access and browser-based stakeholder access without requiring direct database connectivity from client machines.
Integration points: PCS provides the server-side infrastructure that EA GraphLink builds on. EA GraphLink is installed alongside or integrated with PCS: it uses PCS’s repository connectivity to expose the model data through its own API interfaces.
When PCS is required:
Most organizations that have a functioning team Sparx EA deployment already have PCS. The question is whether PCS is current (Sparx Systems releases regular PCS updates) and correctly configured for security and performance.
EA GraphLink is the capability that defines the current era of Sparx EA deployment. It is a connectivity layer that exposes the EA Repository to external systems: specifically, to BI tools and AI platforms: without requiring those systems to know anything about the EA database schema.
EA GraphLink has two interfaces:
Interface A exposes the EA repository as a GraphQL API endpoint. This means Power BI, Tableau, Qlik, and other BI tools can query the architecture model directly: pulling application inventories, capability maps, technology stacks, integration landscapes, and any other structured model data: and display it in live dashboards.
This eliminates one of the most persistent pain points in EA programs: the manual export cycle. Without EA GraphLink, getting model data into a Power BI dashboard requires exporting from EA (CSV, Excel, or custom script), loading into Power BI, and repeating every time the model changes. With EA GraphLink, the Power BI report queries the live model: the dashboard is always current.
For executives and governance boards, this means architecture data is available in the same BI environment as operational and financial data: enabling the kind of integrated decision-making that architecture programs aspire to but rarely achieve.
What GraphQL exposes:
The GraphQL schema reflects the EA repository structure: packages, elements, connectors, diagrams, tagged values. The query capability allows filtering, joining, and aggregating model content. For example:
The richness of the GraphQL output depends directly on the richness of the model: which is why MDG Technology governance is the quality gate for EA GraphLink value.
Interface B exposes the EA repository as a Model Context Protocol (MCP) Server. MCP is the emerging standard (pioneered by Anthropic) for giving AI tools structured access to external data sources. Through the MCP interface, AI assistants: Microsoft Copilot, Salesforce Agentforce, Claude, and others: can query the architecture model as part of a conversation.
What this enables:
An executive asks Microsoft Copilot: “What are the integration dependencies between our Order Management system and our Logistics platform?” Copilot queries the EA repository via EA GraphLink’s MCP Server, retrieves the interface documentation from the architecture model, and returns a structured answer: drawing on the governed architecture repository rather than guessing or searching unstructured documents.
A business analyst asks Claude: “Which applications in our portfolio are vendor-supported, and which are approaching end-of-support?” The AI retrieves the technology lifecycle data from the EA repository and returns a ranked list.
An architect working in Microsoft Fabric asks Copilot to generate an impact analysis for a proposed ERP upgrade: Copilot uses the MCP Server to retrieve the ERP’s upstream and downstream dependencies from the EA model and generates the analysis.
The MDG quality gate:
The quality of AI responses via Interface B depends entirely on the quality of the EA model. A well-governed model: with consistent stereotypes, complete tagged values, documented relationships, and maintained lifecycle data: produces rich, accurate AI responses. A poorly-governed model produces unreliable or incomplete responses. This is why EA GraphLink and MDG Technology governance are always addressed together in Sparx Services’ Connect engagements.
Kernaro AI Hub (General Availability 2026) is a stakeholder-facing platform that provides natural language AI access to the architecture model for non-architect users. It is built on top of EA GraphLink: the AI Hub’s intelligence layer queries the EA repository via EA GraphLink’s MCP Server.
What Kernaro AI Hub does:
Who it is for:
Kernaro AI Hub is designed for the stakeholder audience that currently cannot access the architecture model: business leaders, project managers, technology decision makers, auditors, and board members who need architecture context but will not (and should not) use the Sparx EA desktop client.
What it requires:
Kernaro AI Hub requires EA GraphLink to be implemented and operational. The AI Hub’s query engine talks to EA GraphLink’s MCP Server: which in turn queries the EA Repository. Without EA GraphLink, Kernaro AI Hub cannot function.
This dependency makes the Connect engagement (EA GraphLink implementation) the prerequisite for Kernaro AI Hub deployment.
Kernaro Assist is a separate Kernaro product: an in-EA AI assistant for architects. Unlike Kernaro AI Hub, Kernaro Assist operates inside the Sparx EA desktop client and does not require EA GraphLink.
What Kernaro Assist does:
Who it is for:
Kernaro Assist is for architects who use Sparx EA daily: it accelerates their work without requiring any change to the toolchain or additional infrastructure.
What it requires:
Kernaro Assist requires Sparx EA (Corporate or Ultimate edition). It does not require EA GraphLink or Pro Cloud Server. This makes it independently deployable: an organization can get value from Kernaro Assist immediately, without first implementing the full EA GraphLink stack.
The following describes the integration architecture of the full stack:
At the base: The EA Repository (SQL Server / MySQL / PostgreSQL) is the single source of truth. All model content lives here.
Connecting architects to the repository: The Sparx EA desktop client connects directly to the repository (for local/network deployments) or via Pro Cloud Server (for cloud or remote deployments). PCS manages authentication, authorisation, and connection management.
Connecting BI tools to the repository: EA GraphLink sits alongside PCS, reading from the repository and exposing a GraphQL endpoint (Interface A) on a configured port. Power BI connects to this endpoint using a GraphQL connector or a custom data source. The connection is live: Power BI reports query the current model state on each refresh.
Connecting AI tools to the repository: EA GraphLink also exposes an MCP Server (Interface B) that registers with AI platforms as a tool or data source. Microsoft Copilot, Claude, and other MCP-compatible AI systems discover the MCP Server and can query it during AI sessions.
Connecting stakeholders to the architecture: Kernaro AI Hub is a separate web application that uses EA GraphLink’s MCP interface as its data source. Stakeholders access Kernaro AI Hub via browser: it presents a conversational interface, handles the AI session, queries EA GraphLink for architecture data, and returns formatted responses.
Inside the EA client: Kernaro Assist operates as a Sparx EA Add-In: it reads from the local EA client session and calls Kernaro’s cloud AI service. It does not require the GraphLink stack.
Sparx Systems software (Sparx EA, Pro Cloud Server, EA GraphLink) is purchased directly from Sparx Systems or through authorized resellers. Sparx Services does not resell the software. Our role is to:
Kernaro software (Kernaro AI Hub, Kernaro Assist) is purchased through Kernaro’s commercial arrangements. Sparx Services advises on Kernaro components as part of broader platform engagements.
Sparx Services engagements are professional services: scoped, priced, and delivered by Sparx Services independently of software licenses.
Q: Do I need all six components? A: No. The components are additive: you deploy what you need for your current objectives. Most organizations start with Layers 1–3 (EA, Repository, PCS) and have already done so before engaging Sparx Services. EA GraphLink (Layer 4) is the investment that unlocks BI integration and AI capabilities: it is the most common addition for mature EA programs. Kernaro AI Hub (Layer 5) is the next step for organizations that want to extend architecture access to a broad stakeholder audience. Kernaro Assist (Layer 6) can be added independently of EA GraphLink at any point.
Q: We already have Sparx EA running. What do we need to add EA GraphLink? A: EA GraphLink requires a current version of Pro Cloud Server and a supported repository database. If your PCS and EA are on current versions, EA GraphLink is a relatively straightforward addition: typically a few days of configuration work. If your PCS is significantly out of date, an upgrade may be required first. The Connect engagement begins with a technical assessment of your current stack to confirm readiness and identify any prerequisite work.
Q: What BI tools does EA GraphLink work with? A: EA GraphLink’s Interface A (GraphQL) is compatible with any BI tool that supports GraphQL as a data source or custom connector. Power BI (via the GraphQL connector or a custom data source), Tableau (via web data connector or GraphQL extension), and Qlik are the primary targets. Other tools that support REST or GraphQL connections can also be configured. Power BI is the most common deployment in Sparx Services engagements.
Q: What AI tools does EA GraphLink’s MCP Server work with? A: Interface B supports any MCP-compatible AI platform. Current compatible platforms include Claude (Anthropic), Microsoft Copilot (via MCP integration), Salesforce Agentforce, and Microsoft Fabric. As MCP adoption grows, additional AI platforms are becoming compatible. The MCP Server exposes a defined set of tools (query operations against the EA repository) that MCP-compatible AI platforms can discover and use.
Q: Does MDG Technology governance really affect EA GraphLink quality that much? A: Yes: significantly. EA GraphLink exposes whatever is in the model. If the model has inconsistent stereotypes, missing tagged values, undocumented relationships, and elements named only with abbreviations, that is what the GraphQL API and MCP Server expose. BI dashboards built on a poorly-governed model are unreliable. AI responses drawing on a poorly-governed model are incomplete or misleading. A well-governed model: consistent, complete, with meaningful tagged values and documented relationships: produces dashboard and AI output that is genuinely useful. This is not a marginal difference; it is the difference between an architecture program that delivers value and one that doesn’t.
Q: Can Kernaro AI Hub replace Sparx EA for some users? A: Kernaro AI Hub is not a modeling tool: it cannot create or edit the architecture model. It is a consumption and query interface for stakeholders who need architecture insight but do not need (and should not need) modeling capability. It does not reduce the number of Sparx EA licenses required for your architect team. What it does is extend the reach of the architecture program to a much broader audience: enabling the business value of the architecture to be consumed without requiring Sparx EA training or licenses for every consumer.
Q: What is the difference between Kernaro AI Hub and Kernaro Assist? A: Kernaro AI Hub is a stakeholder-facing platform: a browser-based conversational interface for non-architects to query the architecture model. It requires EA GraphLink. Kernaro Assist is an architect productivity tool: an AI assistant embedded inside the Sparx EA desktop client that helps architects model faster and more accurately. It does not require EA GraphLink. They are complementary products for different users, deployable independently.
Q: Where does Sparx Services fit in the licensing and software supply chain? A: Sparx Systems software (EA, PCS, EA GraphLink) is purchased directly from Sparx Systems. Sparx Services is not a software reseller. Our role is advisory (what to buy and in what quantities), implementation (installing and configuring the stack), and practice development (MDG governance, training, automation). We provide a detailed bill of materials before any software purchase: so you know exactly what you need before committing. The Discover engagement is often the starting point for organizations that want this advisory input before making a platform investment.
For organizations looking to implement EA GraphLink and unlock BI and AI capabilities from their existing Sparx EA investment: or for organizations building a new Sparx EA platform from scratch: the Connect offering is the primary engagement path.
Connect delivers a configured, production-ready Sparx EA platform: including EA GraphLink, Power BI integration, and MDG Technology governance: with the foundation in place for Kernaro AI Hub when the timing is right.
Connect engagements are scoped between $50K and $185K+ depending on the complexity of the existing environment and the scope of BI/AI integration required.
Explore the Connect Offering | Talk to a Sparx Services Architect
Talk to a Sparx Services architect about where your organization is on the journey and what the next stage looks like.