The 2026 Sparx EA ecosystem: what’s new, what’s changed, what matters

The Sparx EA ecosystem moved fast in the first half of 2026. Not everything that moved mattered, but some things did. Here is a practitioner’s read on what actually changed — and what it means for how you build a modern enterprise architecture practice this year.

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 is 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’s GraphQL API for your EA repository is stable and production-ready. If you are 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 is 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 you run — access your architecture model directly. If you are grounding AI in your architecture, the MCP server is usually more straightforward than working with raw GraphQL queries.

Tableau’s MCP support is now generally available. Tableau’s official MCP server reached general availability in late 2025, and it is worth your attention. Tableau’s BI platform can now natively integrate with MCP servers, which means dashboards and analytics can pull directly from your Sparx EA repository through the MCP server. This does not replace EA GraphLink, but it simplifies the integration path for Tableau shops.

Salesforce Agentforce added MCP support. Salesforce’s AI agent builder now speaks MCP, which means you can build Agentforce agents that query your architecture data. If your organization is standardizing on Agentforce for agent development, you can now build agents that understand your system landscape — the same MCP server you already stood up for Copilot does double duty here.

Cursor improved its 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 improved MCP support makes it easier to wire in.

The headline is not any single release. It is that the pieces finally fit together — one MCP server, grounded in one repository, feeding every AI tool your organization already runs.

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’s repository hosting layer) was necessary or whether a file-based repository was sufficient. That debate is over. Any serious EA practice now runs Pro Cloud Server, for a simple reason: EA GraphLink requires it, MCP requires it, and continuous governance requires it. You cannot do modern EA practice on file-based repositories. That 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 quietly. Teams are now treating MDG as a strategic tool, because AI capabilities are only as good as the metadata they work 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’s 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 will not download or learn the desktop tool. It is 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.

Multi-model repository support in Kernaro Assist

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 are watching to see how cleanly it works, because multi-model governance is genuinely hard.

How widely MCP spreads across vendors

Google Gemini and a steady stream of other platforms have added MCP support, and Microsoft Copilot has had it from early on. MCP is becoming the standard protocol for AI tools to reach external data. The open question is reach: if more design, planning, and line-of-business tools support MCP, that changes how architecture data flows into the rest of the organization. We are watching which vendors move quickly — because every new MCP-capable tool is one more place your architecture data can do work without a custom integration.

Whether Agentforce reaches 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), largely because Copilot Studio has more mature MCP support. Salesforce is investing here, but it is not yet clear whether they will reach parity or specialize in a different agent architecture. This matters if your organization is choosing between the two platforms — and either way, grounding the agent in your EA repository is the same MCP move.

What this means for the Ecosystem Explorer

The Ecosystem Explorer — our 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 Explorer is organized by use case, not by tool. So if you are trying to let stakeholders query architecture data without learning Sparx EA, we point you to Kernaro AI Hub and Prolaborate. If you are trying to ground AI tools in architecture data, we point you to EA GraphLink, MCP, and the vendor-specific integrations. If you are trying to build dashboards from architecture data, we point you to Power BI, Tableau, and other BI platforms. It is a free resource for anyone who wants to understand the Sparx ecosystem and how the pieces fit together.

What matters most

Here is the distillation for busy architects.

• Starting or modernizing an EA practice? Begin with Pro Cloud Server, a tight MDG, and Kernaro AI Hub for stakeholder access. That is the foundation; everything else builds on top of it.
• Working day-to-day in Sparx EA? Kernaro Assist is worth adopting. Yes, there is governance overhead up front. The productivity gains are real and measurable.
• Building AI-grounded applications? EA GraphLink and MCP are how you connect to architecture data. Both work — GraphLink is more flexible, MCP is more standardized.
• Making data available to non-architects? Kernaro AI Hub is the product that works. Prolaborate is the supporting player for publishing structured models.
• Pinned to a platform? Microsoft shop: Copilot plus EA GraphLink or MCP. Salesforce shop: Agentforce plus MCP. Tableau shop: Tableau plus MCP is straightforward.

The ecosystem is coherent now in a way it was not eighteen months ago. The pieces fit together, the standards are settling, and you can build a modern EA practice on Sparx tools with confidence that the integrations will work and the architecture will be sound. That was not obviously true in 2024. It is now.

The harder question is not which tools — it is where the leverage is for your team. Connecting the repository is the easy part; deciding which work to augment first is the part that pays off. That is the conversation AI Augmented Architecture is built around, and if you want a structured read on your own readiness, Paralysis to a Plan turns it into a scored, fundable starting point. For the full map of tools and integration patterns referenced above, the Ecosystem Explorer stays current quarterly.