Direct Answer
Connecting Sparx EA to Microsoft Fabric involves seven steps: confirming prerequisites (EA GraphLink deployed, Fabric workspace provisioned, MCP Server enabled), configuring EA GraphLink Interface B for Fabric agent access, setting up a Microsoft Fabric Lakehouse as the data integration layer, configuring the Fabric agent to use the EA MCP Server as a context source, testing with sample queries, building Fabric data flows from EA GraphQL (Interface A) for persistent EA data in Fabric, and connecting Fabric to Power BI for dashboards. The key value of adding Fabric above standalone Power BI: Fabric can enrich EA data by joining it with other organisational data sources (HR, finance, project data) and enable agentic AI data flows that standalone Power BI cannot. The MDG quality gate applies here as for all EA GraphLink connections: Fabric’s intelligence is bounded by the quality of the EA repository it reads from.
Key Takeaways
- Prerequisites: EA GraphLink deployed (both interfaces), Fabric workspace provisioned, Power BI connected.
- EA GraphLink Interface A (GraphQL) feeds EA data into Fabric Lakehouse for persistent, enrichable storage.
- EA GraphLink Interface B (MCP Server) connects to Fabric-embedded Copilot for AI queries about the EA repository.
- Fabric adds: cross-source data enrichment (EA + HR + finance), automated data flows, agentic orchestration.
- Power BI connects to the Fabric Lakehouse for dashboards: more capable than Power BI Direct Query to GraphQL alone.
- MDG quality remains the constraint on output quality throughout.
- This is the full Microsoft EA analytics and intelligence stack: Sparx EA → EA GraphLink → Fabric → Power BI + Copilot.
Prerequisites
Before beginning this integration, confirm:
EA GraphLink is deployed and operational:
- Interface A (GraphQL endpoint) is active and queryable
- Interface B (MCP Server) is configured and accessible
- Pro Cloud Server is running and connected to the Sparx EA repository
- Authentication (API key or OAuth 2.0) is configured for both interfaces
Microsoft Fabric workspace is provisioned:
- A Fabric workspace is created for EA analytics (separate from other Fabric workspaces)
- Appropriate permissions are assigned (Fabric Admin or Workspace Admin for the integration setup)
- Fabric capacity is allocated (F2 or higher recommended for production workloads)
Microsoft 365 Copilot is licensed:
- Copilot M365 licenses are in place for the users and administrators who will configure and use the AI integration
- Copilot is enabled in the M365 Admin Center
MDG governance assessment completed:
- Before connecting Fabric (or any intelligence tool) to the EA repository, confirm MDG governance quality: Are Application Component elements consistently stereotyped? Are Lifecycle Status and Business Owner tagged values populated for the majority of elements? This assessment is performed during the Connect engagement. If significant gaps exist, address MDG governance first.
Step 1: Configure EA GraphLink Interface B for Fabric Agent Access
The Microsoft Fabric agent (Fabric’s agentic AI capability) can connect to external context sources via MCP. Configure the EA MCP Server for Fabric agent access:
In EA GraphLink Interface B configuration:
- Confirm the MCP Server endpoint URL (format:
https://[your-pcs-host]/eagraphlink/mcp) - Generate or retrieve the API key for the Fabric agent integration (separate key from other integrations: for audit traceability)
- Verify that the MCP Server’s capability declaration includes the element types, tagged value dimensions, and query patterns relevant to your intended Fabric agent use cases
- Test the MCP Server endpoint using a tool like Postman or curl: confirm that a capability discovery request returns the expected schema
Fabric connectivity consideration: If the MCP Server is hosted within your corporate network (on-premises or private cloud), the Fabric agent must be able to reach it. Configure the appropriate network path: either expose the MCP Server endpoint via a secure gateway, or use Fabric’s private link capabilities to connect to internal resources.
Step 2: Set Up Microsoft Fabric Lakehouse as the Data Integration Layer
The Fabric Lakehouse stores EA data persistently in Fabric’s Delta Lake format, enabling enrichment with other data sources and more sophisticated analytics than direct GraphQL querying alone.
Creating the EA Lakehouse:
- In the Fabric workspace, create a new Lakehouse: name it “Enterprise Architecture” or equivalent
- The Lakehouse will contain tables for EA repository data: one table per major element type, plus relationship and tagged value tables
Why a Lakehouse rather than direct Power BI connection to GraphQL:
- The Lakehouse enables joining EA data with other organisational data (HR headcount by team, project financial data, ITSM incident data) for cross-source analysis
- Lakehouse tables support full SQL querying: more powerful than GraphQL query syntax for complex analytics
- Fabric’s notebook environment (PySpark/SQL) enables data transformation and enrichment that Power BI’s Power Query cannot match
- The Lakehouse serves as the persistent data layer: EA GraphLink data is loaded on schedule, not queried live on each report load
Step 3: Configure the Fabric Agent to Use the EA MCP Server
Microsoft Fabric includes a Copilot-powered agent capability. Configure the Fabric workspace agent to use the EA MCP Server as an external context source:
In Fabric workspace settings:
- Navigate to Workspace Settings → AI Integration → External Data Sources
- Add a new MCP data source:
– Name: “Sparx EA Repository” – MCP Server URL: the EA GraphLink Interface B endpoint – Authentication: API key (configure the key generated in Step 1)
- Define the agent topic for EA queries: describe what the agent can answer: “Enterprise architecture data including applications, capabilities, technology components, and architecture decisions from the organization’s Sparx EA repository.”
- Save and test the configuration
Validation: Open the Fabric Copilot chat and ask a test question: “What business capabilities does our CRM system support?” If the MCP Server connection is correctly configured, Copilot should query the EA repository and return an answer based on the Realisation relationships in the model.
Step 4: Test with Sample Queries
Before building production data flows and dashboards, validate the integration with a set of representative queries:
Test queries (adjust element names to match your repository content):
Capability coverage: “List all business capabilities with no application coverage in the EA repository.” Portfolio status: “How many application components have a lifecycle status of End-of-Life?” Ownership gaps: “Which application components in the Finance domain have no documented business owner?” Dependency analysis: “Which applications realize the Customer Onboarding capability?” Decision retrieval: “What architecture decisions have been accepted relating to cloud infrastructure?”
Evaluating results:
- Are the answers precise and specific to your repository data? (Not generic answers from training data)
- Do element names, counts, and relationships match what you see in Sparx EA?
- Are any expected elements missing from results? (May indicate MDG governance gaps: elements not correctly stereotyped)
Document the test results. Issues found here indicate either connection problems (if no data is returned) or MDG governance gaps (if data is incomplete or imprecise).
Step 5: Build Fabric Data Flows from EA GraphQL (Interface A)
For persistent EA data in the Fabric Lakehouse, build Fabric data pipelines that query EA GraphLink Interface A and load data into Lakehouse tables.
Fabric Data Pipeline approach:
- In the Fabric workspace, create a new Data Pipeline
- Add a “Copy Data” activity
– Source: GraphQL connector pointing to the EA GraphLink Interface A endpoint – Authentication: API key for Interface A – Query: Write a GraphQL query for the first data set (e.g., all Application Components with their tagged values)
- Configure destination: the EA Lakehouse → target table (e.g.,
ApplicationComponents) - Configure schema mapping: map GraphQL response fields to Lakehouse table columns
- Schedule the pipeline (daily or multiple times daily depending on currency requirements)
Tables to build in the Lakehouse:
For a standard EA analytics configuration:
| Table | GraphQL query scope | Key columns |
|---|---|---|
ApplicationComponents |
All Application Component elements | Name, Stereotype, Domain, LifecycleStatus, BusinessOwner, TechnicalOwner, BusinessCriticality |
Capabilities |
All Capability elements | Name, Domain, MaturityScore, StrategicImportance, ApplicationCoverage, BusinessOwner |
AppToCapability |
Realisation relationships from AppComponent to Capability | SourceName, TargetName, RelationshipType |
TechnologyNodes |
Technology Node elements | Name, Type, Vendor, LifecycleStatus, EOLDate |
ArchitectureDecisions |
ArchitectureDecision elements | ADRID, Status, Domain, DecisionDate, DecisionAuthority |
Incremental loading: For large repositories, configure incremental pipeline runs that load only elements modified since the last run. EA GraphLink supports query filters by modification date.
Step 6: Build Enriched Data in Fabric
This is what Fabric enables beyond standalone Power BI + GraphQL:
Joining EA data with HR data: “sql -- Example: Application ownership enriched with HR manager chain SELECT ac.Name AS ApplicationName, ac.LifecycleStatus, ac.BusinessOwner, hr.ManagerName, hr.Division FROM ApplicationComponents ac LEFT JOIN HRDirectory hr ON ac.BusinessOwner = hr.DisplayName WHERE ac.LifecycleStatus = 'End-of-Life' “
This query (run in a Fabric SQL analytics endpoint) produces a list of End-of-Life applications with their business owners and the owner’s manager: exactly the stakeholder list needed for decommissioning conversations.
Joining EA data with project portfolio data: “sql -- Applications being replaced by active projects SELECT ac.Name AS LegacySystem, ac.BusinessCriticality, pp.ProjectName, pp.PlannedDeliveryDate FROM ApplicationComponents ac JOIN ProjectPortfolio pp ON pp.TargetSystemToReplace = ac.Name WHERE ac.LifecycleStatus IN ('Migrate', 'Eliminate') “
This cross-source analysis requires Fabric: it cannot be built in Power BI alone without pre-joined data.
Step 7: Connect Fabric to Power BI for Dashboards
With EA data in the Fabric Lakehouse, connect Power BI to the Lakehouse for dashboard building:
- Open Power BI Desktop
- Connect to data → Microsoft Fabric → Lakehouse
- Select the EA Lakehouse and the relevant tables
- Build the Power BI data model: create relationships between tables (ApplicationComponents to AppToCapability to Capabilities)
- Build dashboard pages:
– Application Portfolio Heat Map (LifecycleStatus by Domain) – Capability Coverage Matrix (Capabilities vs ApplicationCoverage score) – Technology EOL Timeline (TechnologyNodes by EOLDate) – Architecture Decision Register (status breakdown, trend over time)
- Publish to Power BI Service (in the same Fabric workspace)
- Configure refresh schedule: in the Fabric workspace, the Lakehouse data refreshes via the pipeline schedule; Power BI semantic model refreshes from the Lakehouse on a separate schedule
Optional: Enable Copilot on Power BI dashboards: In Power BI Service, enable Copilot for the published semantic model. This allows M365 users to ask natural-language questions about the dashboard data (“What’s the average maturity score for our Supply Chain capabilities?”) without navigating dashboard pages.
Frequently Asked Questions
Q: What is the minimum Fabric capacity required for this EA integration? A Fabric F2 capacity is the minimum for production workloads. F2 provides sufficient compute for daily pipeline runs, Lakehouse storage for typical EA repository sizes, and Copilot agent access. For organizations with large EA repositories or high-frequency data pipelines, F4 or F8 may be appropriate. Sparx Services advises on Fabric capacity sizing during the Connect engagement.
Q: Can we use this integration without a full Fabric capacity license, using only Power BI? Yes. Power BI can connect directly to EA GraphLink Interface A (GraphQL) without Fabric: using Power BI’s GraphQL connector in Power Query. This is simpler and lower cost, but loses the cross-source enrichment and agentic capabilities that Fabric provides. For organizations that only need EA dashboards (not cross-source analysis or Fabric agent workflows), Power BI direct connection to GraphQL is sufficient. The Fabric integration adds value for organizations that want to combine EA data with other enterprise data sources.
Q: Does the Fabric agent access the MCP Server in real time, or does it use the Lakehouse data? The Fabric agent accesses the EA MCP Server in real time for natural-language query responses: it queries current repository data via the MCP protocol, not the Lakehouse. The Lakehouse is used by Power BI dashboards (for structured analytics) and by Fabric SQL notebooks (for cross-source enrichment queries). The two access patterns coexist: real-time AI querying via MCP; structured analytics and enrichment via Lakehouse.
Q: How often should EA data be refreshed in the Fabric Lakehouse? Daily refresh is appropriate for most EA analytics use cases: architecture data changes on a timescale of days or weeks, not minutes. For executive dashboards where yesterday’s data is current enough, a daily pipeline run overnight is the right configuration. If your EA team makes significant updates during the day and stakeholders need the same-day updates reflected in dashboards, configure multiple pipeline runs (every 4–6 hours). Real-time AI querying (via MCP) is always current regardless of Lakehouse refresh schedule.
Q: What are the security implications of EA data being stored in a Fabric Lakehouse? EA data in a Fabric Lakehouse is subject to Microsoft Fabric’s security model: workspace role-based access, Azure Active Directory integration, and Microsoft Purview data governance. Architects who model in Sparx EA may not have access to the Fabric Lakehouse; Lakehouse access is managed separately. For sensitive architecture data (security architecture details, systems holding sensitive data), consider whether all EA data should be ingested into Fabric, or whether a subset schema (excluding sensitive packages) should be defined for the Fabric integration. This is addressed in Connect engagement scoping.
Q: Can Fabric’s automated machine learning features be applied to EA data? Yes. Once EA data is in the Fabric Lakehouse, Fabric’s AutoML and data science notebook capabilities can be applied. Practical use cases include: predicting application obsolescence risk based on age, vendor EOL patterns, and technical health scores; identifying capability investment priority based on historical patterns; clustering applications by similarity for rationalization analysis. These analytics go beyond standard BI dashboards and represent an advanced use case for mature EA programs with sufficient data history in the Lakehouse.
Ready to Connect Sparx EA to Microsoft Fabric?
Sparx Services’ Connect engagement includes EA GraphLink deployment, Fabric Lakehouse configuration, data pipeline setup, and initial Power BI dashboard delivery.
For organizations on the Microsoft stack, this is the full EA analytics and intelligence path: Sparx EA → EA GraphLink → Fabric → Power BI + Copilot.
