The 2026 Sparx EA ecosystem: what each new piece lets an architect do

Several new tools arrived around Sparx EA in 2026. Reading them as a product list is less useful than asking what each one lets an architect, or the people an architect serves, get done. This review takes them in that order, then looks at what has changed in how teams use the tools they already had.

Stakeholders can ask the model questions

Kernaro AI Hub launched in January 2026. It is a Sparx Systems web application that lets people who do not use Sparx EA ask questions of the architecture model in plain language. It includes a read-only MCP server and a read-only GraphQL API, so other AI tools in the organization can be grounded in the same model. Kernaro is bring-your-own LLM: your organization supplies the keys and pays the provider for usage.

For the architect, the change is in where their week goes. Routine questions about what connects to what can be answered from the model directly, so less time goes to relaying facts. The skill it asks is keeping the model current and well structured, because every answer the Hub gives is only as good as the repository behind it.

Architects can direct AI to do the modeling

AI Power Tools for Sparx EA launched in April 2026. Since October 2026 it is free for every Sparx user and published by NovoCircle. It gives an AI full read and write access to the live repository from the architect’s own Windows machine, with Enterprise Architect on the same machine. For Claude, the Claude Desktop extension is the MCP server and the ai-power-tools plugin carries the skills, the written procedures the AI follows; both are needed. GitHub Copilot in VS Code gets the same server through a VS Code extension, and VS Code loads the same skills.

This is the piece that moves modeling itself from manual to AI-augmented. The architect decides what the model should say, and the AI builds the diagrams, completes the mappings and checks the result against the team’s rules. The skill it asks is stating intent precisely and reviewing what comes back. The server is local to the architect’s machine, so it serves the person doing the modeling, not a team of readers.

Architects get help inside the EA client

Kernaro Assist launched in May 2026. It is a Sparx Systems AI assistant that works inside the Enterprise Architect client, also bring-your-own LLM. It suits architects who want AI help without leaving the tool they already model in. What it produces depends heavily on the modeling language behind the repository, which is why MDG work has moved up the agenda (more on that below).

Stakeholders get AI on the published model

Prolaborate 6 with AI. Sparx Systems describes it as generating descriptions, summarizing discussions and diagrams, rephrasing, and answering questions about the model. It is inactive by default and needs a Kernaro AI Hub license and your own LLM. For an architect who already publishes to business stakeholders through Prolaborate, it adds help with the writing and explaining around the published model.

Architecture data reaches reporting and the Microsoft estate

Leaders want architecture facts in the dashboards and assistants they already use. We can build a reporting database on top of your Sparx EA repository. If you host the repository yourself, we build the reporting views in the same database and you read from them live. If you cannot reach the database, or it is hosted elsewhere, and you need the reporting data in something like Snowflake or Microsoft Fabric, we help you set it up so you refresh it from your EA model on the schedule you define.

From there we can help you get the data where people work. For Power BI, the Kernaro Connector for Power BI, an extension to Kernaro AI Hub, is the preferred route when your organization has the Hub; otherwise, the Power BI model that AI Power Tools generates from the reporting database. Microsoft 365 Copilot answers today from a published Power BI semantic model, or, with Kernaro AI Hub, through the Hub’s read-only API via Copilot Studio or a Microsoft 365 connector. For Microsoft Fabric, the reporting database can be mirrored into Fabric, and a data warehouse such as Snowflake can be fed from the reporting database or from Parquet files. The reporting and Power BI page covers this in more depth.

Other platforms that speak MCP

MCP is an open protocol, and support for it is spreading across BI platforms, agent builders and code editors. If your platform can call an MCP server, it can in principle be pointed at Kernaro AI Hub’s read-only server. Whether and how a given platform supports MCP is that vendor’s decision, so confirm it with them before you plan around it. The AI Power Tools server is different: it runs on the architect’s own Windows machine and is not a shared endpoint for other platforms. Its documented clients are the Claude desktop app and GitHub Copilot in VS Code; Cursor is not a documented surface.

One governed repository, more than one standard route into it, and a choice of AI tools on top. Which tool fits depends on the job.

What has changed in how teams use what they already had

Shared, server-side access needs a server-side foundation. Pro Cloud Server, Sparx’s repository hosting layer, comes into the conversation as soon as several teams need to reach the same repository through a web product. Confirm with Sparx Systems what your Kernaro AI Hub deployment requires. The desktop routes are different. AI Power Tools and Kernaro Assist work from the architect’s own machine with EA open, so an individual architect can start there without a server project.

MDG Technology matters more. The modeling language definition used to be treated as technical plumbing that an administrator looked after. AI work raises its importance, because an AI follows the language it is given. A loose MDG produces loose AI output, and a tight MDG with clear rules produces content an architect can check quickly. The architect designs the governance intent, and AI applies it and checks the model against it.

Prolaborate is the publication route for business stakeholders. Architects model in Sparx EA and publish through Prolaborate for stakeholders who will not learn the desktop tool. That pattern was already common, and the AI additions in Prolaborate 6 build on it.

What Sparx Services is watching

Architectures that span more than one repository

Federated models, where several teams contribute to one logical architecture, matter for distributed enterprises and for organizations that inherited more than one EA practice. How the Sparx tooling handles that is a question for Sparx Systems, and we are watching what they say, because multi-model governance is hard.

How widely MCP spreads

If more design, planning and line-of-business tools support MCP, architecture data can be used in more places without a custom integration. Support is each vendor’s decision, so check it product by product.

Which agent platform your organization standardizes on

Microsoft and Salesforce both offer agent builders, and each decides what its agents can connect to. If your organization is choosing between them, put the MCP question to each vendor directly. Where an agent can call an MCP server, grounding it in your EA repository works the same way.

Where to start, by job

• Modernizing how the team models: start with a clear MDG, then let architects direct AI Power Tools on their own machines, or try Kernaro Assist inside the EA client on a bounded part of the model.
• Letting non-architects get answers: Kernaro AI Hub is built for it, with Prolaborate for publishing the structured model.
• Grounding other AI tools in architecture data: Kernaro AI Hub’s read-only MCP server or API, where the platform supports it.
• Getting architecture data into dashboards and Microsoft 365 Copilot: a reporting database, then the Power BI and Microsoft routes above.

The harder question is which work to augment first for your team. That is the conversation AI-Augmented Architecture is built around. If you want a structured read on your own starting point, Paralysis to a Plan turns it into a scored plan you can fund. For a plain comparison of the AI options, including where each one stops, see AI tools for Sparx EA. The Ecosystem Explorer maps the wider set of integration points by the job you are trying to do.