The Sparx EA ecosystem moved fast in the first half of 2026. Not everything that moved mattered, but some things did. Here’s what you need to know.
What’s new since January
Kernaro AI Hub went GA. This was the big one. The external web application that lets non-architects query your architecture model in plain English through a Copilot-like interface moved from limited availability to general availability. If you haven’t kicked the tires on Kernaro AI Hub yet, now’s the time. This is the product that makes your architecture data available to stakeholders without teaching them to use Sparx EA.
EA GraphLink matured. Sparx Systems’ GraphQL API for your EA repository is stable and production-ready. If you’re building dashboards, doing custom BI integration, or grounding AI tools in architecture data, EA GraphLink is how you do it. The query performance is solid. The schema is stable. It’s no longer beta.
EA GraphLink got an MCP server. The Model Context Protocol server for EA GraphLink is now mature enough for production use. This is the mechanism that lets AI assistants (Copilot, Claude, Agentforce, whatever) access your architecture model directly. If you’re grounding AI in your architecture, the MCP server is usually more straightforward than direct GraphQL queries.
Tableau announced MCP support. This is in public preview but worth watching. Tableau’s BI platform can now natively integrate with MCP servers. For EA, this means dashboards and analytics can pull directly from your Sparx EA repository through the MCP server. This doesn’t replace EA GraphQL, but it simplifies the integration path for Tableau shops.
MuleSoft Fabric announced MCP capability for Agentforce. Salesforce’s AI agent builder now supports MCP, which means you can build Agentforce agents that query your architecture data. If your organization is standardizing on Agentforce for AI agent development, you can now build agents that understand your system landscape.
Cursor improved MCP tooling. The AI-native code editor Cursor got better at discovering and configuring MCP servers, reducing the friction of hooking up external data sources. For architects and analysts writing scripts or doing custom integrations, Cursor is a useful tool, and the MCP support makes it easier.
What’s changed in how teams use what they already had
Pro Cloud Server is now table stakes. Two years ago, teams debated whether Pro Cloud Server (Sparx Systems’ on-premises repository hosting) was necessary or if Project Sparx (file-based collaboration) was sufficient. That debate is over. Any serious EA practice now runs Pro Cloud Server. The reason: EA GraphLink requires it, MCP requires it, continuous governance requires it. You can’t do modern EA practice on file-based repositories. This has become clear.
MDG Technology is getting strategic attention. The metamodel definition system used to be thought of as technical infrastructure, something the EA admin managed. Teams are now treating MDG as a strategic tool. The reason: AI tools are only as good as the metadata they’re working with. If your MDG is sloppy, Kernaro Assist generates sloppy content. If your MDG is tight, Assist generates tight content. MDG governance went from an operational concern to a strategic one.
Prolaborate adoption is accelerating. Sparx Systems’ lightweight web interface for viewing and collaborating on architecture models is showing up in more organizations. The pattern is clear: architects work in Sparx EA itself, but use Prolaborate as the publication mechanism for business stakeholders who won’t download or learn Sparx EA. It’s becoming the bridge between the EA repository and the broader organization.
What Sparx Services is watching
Three things are on our radar as we move into the second half of 2026.
Kernaro Assist’s roadmap around multi-model repositories. Kernaro Assist currently works with single Sparx EA repositories. The roadmap includes support for architectures that span multiple repositories (federated models, multiple organizational teams contributing to a single logical architecture). This matters for distributed enterprises and for organizations with inherited EA practices. We’re watching to see how cleanly it works. Multi-model governance is hard.
MCP adoption across other vendors. Google Gemini announced MCP support in preview. Microsoft Copilot has had it from the start. This is becoming the standard protocol for AI tools to access external data. The question is: how widely will it spread? If Adobe Creative Suite supports MCP, that changes how architecture data flows into design tools. If SAP SuccessFactors supports MCP, that changes how architecture data influences HR planning. We’re watching to see which vendors move quickly.
Whether Agentforce’s MCP implementation reaches feature parity with Copilot Studio. Right now, more enterprise AI agent development is happening in Copilot Studio (Microsoft’s agent builder) than in Agentforce (Salesforce’s). The reason: Copilot Studio has more mature MCP support. Salesforce is investing in this, but it’s not clear yet whether they’ll reach parity or whether they’ll specialize in a different agent architecture. This matters if your organization is choosing between these platforms.
What this means for the Ecosystem Directory
The Ecosystem Directory (Sparx Services’ curated guide to integration points and tools in the Sparx EA world) has been updated for 2026. The biggest changes:
- Kernaro Assist moved from the “emerging” section to “foundational”
- EA GraphLink moved to “foundational”
- MCP integrations now have their own section
- Prolaborate appears more prominently as a stakeholder-facing publishing layer
- Pro Cloud Server is now explicitly required, not optional
The Directory is organized by use case, not by tool. So if you’re trying to “let stakeholders query architecture data without learning Sparx EA,” we point you to Kernaro AI Hub and Prolaborate. If you’re trying to “ground AI tools in architecture data,” we point you to EA GraphLink, MCP, and the vendor-specific integrations. If you’re trying to “build dashboards from architecture data,” we point you to Power BI integration, Tableau, and other BI platforms.
The Directory is maintained and updated quarterly. It’s a free resource for anyone who wants to understand the Sparx ecosystem and how pieces fit together.
What matters most
Here’s the distillation for busy architects:
If you’re starting an EA practice or modernizing an existing one, start with Pro Cloud Server, a tight MDG, and Kernaro AI Hub for stakeholder access. That’s the foundation. Everything else builds on top of it.
If you’re an architect working in Sparx EA, Kernaro Assist is worth adopting. Yes, there’s governance overhead initially. The productivity gains are real and measurable.
If you’re building AI-grounded applications (AI agents, Copilot integrations, custom tools), EA GraphLink and MCP are how you connect to architecture data. Both work. EA GraphLink is more flexible. MCP is more standardized.
If you’re trying to make architecture data available to non-architects, Kernaro AI Hub is the product that works. Prolaborate is a supporting player for publishing structured models.
If you’re working in a Microsoft shop, Copilot + EA GraphLink or MCP is the path. If you’re in a Salesforce shop, Agentforce + MCP is the path. If you’re doing Tableau, Tableau + EA GraphQL is straightforward.
The ecosystem is coherent now in a way it wasn’t eighteen months ago. The pieces fit together. The standards are settling. You can build a modern EA practice using Sparx tools and have confidence that the integrations will work and the architecture will be sound.
That wasn’t obviously true in 2024. It is now.
For more detail on any of these tools, integration patterns, and use cases, check out the updated Ecosystem Directory. And if you’re exploring how to modernize your EA practice, The Journey is the place to start the conversation.
