AI Integrations
EA GraphLink is the connectivity and transformation layer that sits between your EA Repository and every downstream tool: Power BI, Copilot, Fabric, Claude, Tableau, Agentforce, and more. It connects directly to the repository database (not through the EA client), transforms the physical schema into a logical architecture model using your MDG Technology definition, and exposes two interfaces: a GraphQL API for BI dashboards and an MCP Server for AI tools and data integration platforms.
You deploy EA GraphLink once. All downstream tools connect to it. No separate deployments for different tools. No manual exports. No periodic snapshots. One live architecture data source, available everywhere.
This is the foundational deployment. Every integration guide you read next: Power BI, Copilot, Tableau, Agentforce, Claude: assumes EA GraphLink is already deployed and configured.
Before EA GraphLink, architecture data lives in Sparx EA. It’s rigorous, well-modeled, and completely invisible to 95% of your organization.
Stakeholders and executives can’t query it. BI teams can’t consume it. AI tools can’t reference it. Integration platforms can’t synchronize against it. The data exists. It’s just locked.
What happens instead:
This is expensive. Architects spend 15–30% of their time answering questions that architecture data could answer automatically. BI teams duplicate effort. Decisions lack current context. Governance fragments across systems.
| Activity | Before EA GraphLink | After EA GraphLink |
|---|---|---|
| Executive dashboards | Manual Excel exports every month, PowerPoint briefings, snapshot data | Live Power BI or Tableau dashboards, architecture updates reflected automatically |
| Stakeholder questions | Email to architect, 2–3 day turnaround, mostly wrong because context is stale | Query in Power BI self-service, or ask Copilot/Agentforce, answered immediately with live data |
| BI team data source | Separate repository or Excel dump, synchronized quarterly | Live GraphQL API connection, automatic updates, single source of truth |
| Integration platforms | Each platform maintains its own architecture metadata | MCP Server connection, all platforms reference the same authoritative source |
| AI context for architects | Tools have no architecture context | Claude, Cursor, Copilot reference live architecture while architects work |
| Architect time | 20–30% spent answering ad-hoc questions | 20–30% freed for strategic architecture work |
EA Repository (SQL Server / MySQL / PostgreSQL)
↓
EA GraphLink
/ \
GraphQL API MCP Server
↓ ↓
Power BI Claude
Tableau Copilot
Cursor
Agentforce
Fabric
MuleSoft
Azure OpenAI
EA GraphLink connects directly to the EA Repository database. It does not go through the EA client or Pro Cloud Server (PCS). This is important:
EA GraphLink requires:
This is the critical step. The EA Repository physical schema is complex: dozens of tables, referential relationships, convention-based patterns. It’s not designed to be queried directly by BI tools or AI systems. It’s designed to be edited by the EA client.
The MDG Technology definition describes the logical architecture model: “This element type is a System. That relationship is a Dependency. This tagged value is the Criticality rating.”
EA GraphLink reads the physical schema and transforms it using the MDG definition. Out comes a clean, structured logical model: elements, relationships, attributes, properties: all described in plain architecture language.
Example:
t_object table, object_type = ‘System’, stereotype = ‘EAI System’, tagged value ea_criticality_score = 8This transformation is what makes the data useful. A well-governed MDG produces rich outputs. A fragmented MDG produces confused outputs.
Once the logical model is built, EA GraphLink exposes two interfaces:
| Interface | Protocol | Connected To | Primary Use |
|---|---|---|---|
| GraphQL | HTTP/REST | Power BI, Tableau | BI dashboards, structured reporting |
| MCP | MCP Standard | Claude, Copilot, Cursor, Gemini, Agentforce, Fabric, etc. | AI context, conversational query, data integration |
Both interfaces expose the same underlying data, transformed by the same MDG. You choose which tools to connect based on your organization’s needs: or connect both for full coverage.
Here’s the hard truth: EA GraphLink is only as useful as your MDG Technology definition.
If your MDG is well-governed:
Then EA GraphLink transforms that into rich, specific, queryable architecture intelligence. AI tools get meaningful context. BI dashboards show clear patterns. Data integration is reliable.
If your MDG is fragmented:
Then EA GraphLink transforms that fragmentation into confident, confidently wrong answers at scale. No error messages. No warnings. Just plausible-looking outputs that don’t mean what you think they mean. Executives make decisions against misleading dashboards. AI tools give architecturally nonsensical advice. Data integration fails silently.
Well-Governed MDG
Output quality: Rich, specific, actionable. “Show me all systems supporting Customer Onboarding” returns exactly what you need.
Partially Governed MDG
Output quality: Mixed. Some queries work perfectly, others are confusing. “Show me systems supporting Customer Onboarding” works if Capability relationships are consistent, fails if one domain uses Realizes and another uses Traces.
Informal or Ad-Hoc MDG
Output quality: Confused or misleading. “Show me systems supporting Customer Onboarding” returns a mix of Realizes, Traces, and undocumented custom relationships that no one can explain consistently. BI dashboards show contradictory numbers. AI tools produce plausible-sounding nonsense.
If your MDG is in poor shape, do not start a Connect engagement. Start with:
The sequence is important. Connect on top of a fragmented MDG wastes budget. Deploy first, then Connect.
Power BI and Tableau dashboards that update automatically as architects model changes. No manual exports, no monthly refresh cycles. Executives see current architecture state.
Business stakeholders ask architecture questions in Copilot or Agentforce without learning the EA client. “Which systems handle payment processing? What’s the roadmap for deprecating the legacy middleware?”
Engineers in Cursor, architects in Claude, data engineers in Azure OpenAI: all have live architecture context while they work, without leaving their tools.
Fabric and MuleSoft teams consume architecture data as a first-class data source. Governance, roadmaps, and system relationships flow into data integration workflows automatically.
AI tools can reason about architecture at scale. “Analyze all systems with external integrations and flag compliance risks”: answered by Claude in seconds with live data.
Kernaro AI Hub provides a browser-based interface for executives and business stakeholders to ask architecture questions in plain English. “What’s our cloud adoption status?” “Which legacy systems are we retiring?” “Show me the roadmap for this capability.”
| Persona | How They Benefit | Which Integration |
|---|---|---|
| Executive / C-suite | Live architecture dashboards, answerable questions via Copilot or Kernaro | Power BI, Copilot, Kernaro Hub |
| Business Stakeholder | Self-service architecture queries, natural language interface, no learning curve | Kernaro Hub, Copilot, Tableau |
| Architecture Manager | Visibility into portfolio, live reporting, reduced ad-hoc question volume | Power BI, Tableau, Discover/Deploy work |
| Solution Architect | Live architecture context in Cursor while designing, Claude analysis of design tradeoffs | Cursor, Claude |
| Data Architect | Architecture as data source in Fabric or MuleSoft, live governance metadata | Fabric, MuleSoft |
| EA Architect | Live context while modeling in EA, AI-assisted analysis, audit trail for governance | Claude, Cursor, Power BI |
| Integration Architect | Architecture-driven integration patterns, live system and relationship data | MuleSoft, Fabric |
| Compliance/Governance | Audit-ready architecture data, live traceability, automated compliance checks | Fabric, data platforms |
Quarterly architecture reports, monthly executive briefing updates, ad-hoc exports to PowerPoint: all replaced by live dashboards that update automatically.
One EA client license = one person seeing the models. One EA GraphLink deployment = unlimited stakeholder visibility via Power BI, Tableau, Copilot, and AI tools. Architecture becomes an enterprise asset, not an architect’s tool.
Not locked to Power BI. Not locked to Copilot. Not locked to Salesforce. Deploy EA GraphLink and connect Power BI, Tableau, Claude, Copilot, Agentforce simultaneously. If you add new tools later, they connect to the same deployment.
MCP is the emerging standard for AI tool integrations. Sparx Systems chose to support it rather than proprietary APIs. Your investment isn’t locked to one AI vendor.
Live connection to the repository, not periodic exports or batch updates. When architects model a change, stakeholders see it immediately.
Stakeholders answer their own questions in dashboards or AI tools. Architects stop context-switching for “quick” architecture questions. Time freed for strategic design.
EA GraphLink is a separate Sparx Systems software product, not included with EA. You must purchase a license directly from Sparx Systems. Sparx Services provides the bill of materials.
You must be on a supported version of Sparx Systems Architecture Platform. Older versions are not compatible. If you’re on legacy versions, upgrade work may be required.
If your MDG is fragmented or informal, EA GraphLink won’t help: it will surface confusion at scale. You must complete Discover (assessment) or Deploy (governance) before Connect. There are no exceptions to this.
Connecting EA GraphLink to Power BI, Copilot, Fabric, etc. requires integration work. Sparx Services handles this in the Connect engagement, but the effort is real: 8–16 weeks depending on scope and number of integrations.
Before connecting to cloud AI tools (Copilot, Agentforce, Claude), your organization’s data governance team should review. This is a compliance activity, not a technical one: but it’s necessary before moving live.
EA GraphLink deployment may require brief repository downtime. This is coordinated during maintenance windows.
| Requirement | Details |
|---|---|
| EA Repository Database | SQL Server, MySQL, or PostgreSQL on supported versions. Must be network-accessible to EA GraphLink deployment. |
| EA Repository Version | Sparx Systems Architecture Platform version on current or recent support line. Check with Sparx Systems for compatibility matrix. |
| EA GraphLink License | Purchased directly from Sparx Systems (not included with EA). Sparx Services provides bill of materials during scoping. |
| Network Access | EA GraphLink server must connect to repository database. Firewall rules, security groups, VPC rules configured appropriately. |
| Downstream Tool Licenses | Power BI (Microsoft), Tableau (Salesforce), Copilot (Microsoft), Agentforce (Salesforce), Claude (Anthropic), etc. depending on which integrations you want. |
| Requirement | Why It Matters |
|---|---|
| MDG Assessment | Discover or Deploy engagement completes before Connect. You can’t skip this. |
| Data Governance Review | Before connecting to cloud AI, data governance and security teams review and approve. |
| Access Control Planning | Who can query architecture in BI tools? Which stakeholders can ask questions in Copilot? Define roles and permissions. |
| Change Management | When stakeholders suddenly have access to live architecture data, change management is needed. Sparx Services supports this. |
When EA GraphLink is live, you stop doing:
This adds up. Architect hours freed per week × hourly rate × 52 weeks = annual savings baseline. Plus: reduced BI team duplicate effort, faster executive decision-making, better architecture compliance visibility.
Use this framework:
Baseline: Architect hours per week answering ad-hoc stakeholder questions × hourly rate × 50–70% self-service adoption rate (with dashboards and AI tools) = annual hours freed.
Example: 3 architects × 5 hours/week × $150/hour × 52 weeks × 60% self-service adoption = $140,400/year freed.
Plus: Portfolio reporting cycle time × hours per cycle × number of cycles per year × hours freed per cycle.
Example: Monthly portfolio report × 12 hours/month × 12 months × 75% time saved = $54,000/year.
Plus: BI team duplicate effort reduction (if you have separate architecture data in your data warehouse).
Total annual value baseline: $150K–$300K+ depending on organization size, architecture complexity, and BI maturity.
Engagement cost: $50K–$185K+ (Connect service).
ROI: Typically positive within 6–18 months.
Export architecture data from EA to Excel/CSV. Import into Power BI or Tableau. Refresh monthly. Works if your architecture complexity is low and stakeholder demand is light. Doesn’t scale.
Some organizations write SQL directly against the EA Repository. This works for technical people, but bypasses the MDG (the logical model): you’re querying the physical schema directly. High technical risk, poor maintainability.
Sparx Systems’ browser-based EA client. Stakeholders get browser-based read access without learning the full EA client. Lower cost than EA GraphLink. Useful for some use cases. But doesn’t integrate with BI tools, AI tools, or data platforms. Not a replacement for EA GraphLink, a complement to it.
No equivalent. There’s no open-source tool that does what EA GraphLink does: MDG-aware transformation of EA Repository schema into GraphQL + MCP interfaces. If you’re committed to open-source, you’re committing to building or forking a transformation layer yourself. Not recommended.
EA GraphLink is a connectivity and transformation layer. Pro Cloud Server (PCS) is a cloud collaboration platform. They serve different purposes:
You can use one, the other, both, or neither: they’re independent. EA GraphLink works with both on-premises and cloud-hosted repositories.
No. EA GraphLink requires a compatible version of Sparx Systems Architecture Platform. Older versions (more than 2–3 major versions back) are not supported. Check the compatibility matrix with Sparx Systems. If you’re on an older version, an upgrade may be required.
SQL Server (supported versions), MySQL (supported versions), and PostgreSQL (supported versions). Other databases are not supported.
No. EA GraphLink connects directly to the repository database. EA client machines don’t need any additional software, plugins, or configuration for EA GraphLink to work. Integration work is on the EA GraphLink server and the downstream tool (Power BI, Copilot, etc.).
GraphQL is designed for BI tools like Power BI and Tableau. It returns structured data optimized for dashboards, charts, and reports. MCP is designed for AI tools and data integration. It returns architecture as queryable resources and context, suitable for AI reasoning and downstream system integration. Both interfaces access the same underlying data; the difference is the protocol and the use case.
Yes. One EA GraphLink deployment serves both. Power BI and Copilot connect to the GraphQL and MCP interfaces, respectively. Tableau and Agentforce connect simultaneously. You’re not locked to one ecosystem.
Completely. If your MDG is well-governed, EA GraphLink produces rich, specific outputs. Queries return meaningful results. If your MDG is fragmented, EA GraphLink transforms that fragmentation into confident, confidently wrong answers. It’s not a software limitation: it’s a data quality issue. Garbage in, garbage out, but without an error message to warn you.
Directly from Sparx Systems. Sparx Services provides the software bill of materials as part of the Connect engagement. You initiate the purchase with Sparx Systems sales. Sparx Services is not a reseller.
No. EA GraphLink is a separate software product. It works with both on-premises and cloud-hosted repositories, but it is not part of the cloud platform subscription. You must purchase an EA GraphLink license separately.
No. Do not deploy EA GraphLink on a fragmented MDG. Instead, start with Deploy ($30K–$130K) to establish governance and ensure consistent MDG application across your repository. Once Deploy is complete and your MDG is in good shape, then proceed to Connect. This sequencing ensures your investment in EA GraphLink produces high-value outputs.
EA GraphLink is the foundation. Everything downstream: Power BI dashboards, Copilot integration, AI context for architects, data platform connectivity: depends on it being deployed and configured correctly.
Before you start: Assess your MDG readiness with Discover ($25K–$75K). If governance work is needed, add Deploy ($30K–$130K). Once your foundation is solid, deploy EA GraphLink as part of Connect ($50K–$185K+).
Three next steps:
Then explore the downstream integration guides:
Talk to a Sparx Services architect about where your organization is on the journey and what the next stage looks like.