AI Power Tools for Sparx EA Docs Release 3.6.4

What you can do with it

The areas below follow the four things AI does for an architect: modeling (the first three areas), analysis, governance (validation, modeling languages and change management) and stakeholder engagement (reporting and Prolaborate). The AI Augmentation Scenarios are grouped the same way. Each area says what you can do, which tool and skills are involved, and where to read more. Operation and skill names appear in code font because they show in the assistant's transcript. Every operation is explained in MCP server: tools and operations and every skill in Skills.

AI Power Tools works through four capabilities: the MCP server, computer use, local code execution and browser control. Each area below lists the capabilities it uses. The MCP server carries all EA work. Computer use covers what the server cannot see or reach in EA's own window. Local code execution runs the code engines that compute fixed answers. Browser control drives web applications. See Capabilities and surfaces for what each enables and what happens without it.

Modeling

Create, read, update, move and delete packages, elements, connectors, class attributes and operations, and tagged values. Create content in the vocabulary of a loaded modeling language. Create hundreds of elements or connectors in one request. Copy a package as new, with its diagrams. Set up click-through (composite) diagrams that navigate down a hierarchy.

Diagrams

Create diagrams and place objects and connectors. Lay out with EA's own layout engine, or compose coordinates from a layout grammar so the diagram reads as deliberate. AI Power Tools defines 11 layout grammars: 9 are composed, one uses EA's own layout and is tidied afterward, and one is EA-semantic, where the diagram type fixes the arrangement and the AI says so instead of composing. Tidy (align and resize), set per-diagram element appearance, connector routing and display settings, collapse parallel connectors into summary views, draw elements as images or as vendor icons taken from EA's shipped icon libraries, check the placed geometry, and export PNG, SVG and editable Visio. Read a diagram, or every diagram in a package, as a manifest that another tool can redraw it from, and write a plain-language description into each diagram so that a reader without EA, a screen reader or an AI can tell what it shows.

Import and export

Export a package as XMI, in any of six XMI versions, with a read-back check that the elements are really in the file. Import XMI as a copy. Generate an EA document from a template. Export diagrams as PNG, SVG or editable Visio. Read a diagram or a package as a manifest to redraw it in another tool, and give diagrams plain-language descriptions, audited and kept current without losing what people wrote. See Diagram descriptions and Reading a diagram or a package as a manifest.

Analysis and traceability

Summarize stereotype, tagged-value and connector-pattern usage. Walk the neighborhood around an element. Find every shortest path between two elements. Build a rooted traceability tree over the relationship types you choose. Render an element in business language, and a diagram in words. Query the audit trail. Run direct database queries with a check that catches wrong column names before EA sees them.

Validation and governance

Run a conformance ruleset against the whole model or one package: required tagged values, connector cardinality, allowed stereotypes at each end of a connector. Fetch a ruleset from a secure web address. Author a ruleset for your own language. Run the shipped ArchiMate 3.1 ruleset. Roll up a portfolio by stereotype and tagged value. Find orphan and ungoverned elements and fix decay with safety checks before anything is deleted.

Modeling languages (MDG Technologies)

Find out which technologies are registered, which are embedded in the model file, and which are actually loaded in the running EA session. These are three different states. List available modeling languages, read a loaded technology's real stereotypes and tagged values, read and write MDG Technology files, install one, export a profile package as a versioned UML profile, and look up display terms for business-language output.

Change management

Snapshot a package or the whole model as a baseline, list baselines, compare a baseline with the model and get a structured difference, restore a package from a baseline, see what changed in a date range and who changed it, and copy a package as new.

Restoring from a baseline discards later changes to that package. Deletes made through EA's automation interface cannot be undone, so take a baseline first.

Reporting database and BI

Build the business layer for a repository: a set of tables named and shaped by the modeling technology that governs the model, over EA's own data. It holds one entity table per element stereotype, one connector table per combination of source stereotype, connector stereotype and target stereotype that the technology allows, tag-row tables only where a column cannot hold the data, and a Table Directory that indexes them. The layer is the same whichever way it is delivered: same table names, columns, types and rows. There are three ways to deliver it, and the facts of your estate decide which are open to you:

  • Views in the EA repository's own database. The business layer is built as SQL views, in a schema of their own, inside the SQL Server database that holds the repository. EA's tables are never altered. The data is live, and Power BI reads it in DirectQuery or Import. This needs direct access to that database and the right to create views in it.
  • A separate reporting database. AI Power Tools reads a scoped copy of EA's tables through Enterprise Architect, so no direct database access is needed. It loads the copy into a separate database, on a SQL Server you run or in a SQL database in Microsoft Fabric, and builds the same views there. The result is a snapshot, rebuilt in full on every refresh, and Power BI imports it.
  • Parquet files, with no database. The same scoped copy is read through Enterprise Architect, and the business tables are written as one Parquet file per table in a local folder, together with the Power BI project that reads them. The result is a snapshot, rewritten on every refresh, and Power BI imports it.

You set the scope (the root packages to include and the packages to leave out) and make the inclusion choice (what the layer includes beyond what the technology defines). Both are saved in the reporting profile, a file that every build, check and refresh reuses. Before a build, the governance gate lists the in-scope elements that carry no stereotype at all, and so would be in no business table, and you decide for each one. For a reporting database or Parquet files, the reporting preflight then reports what would be kept and anything that blocks the build, and writes nothing. After a build, the alignment check measures the layer against the technology's metamodel, and the value check compares every cell with an independent read through EA. A reporting database or Parquet build keeps the extract it loaded, so replay can rebuild that build's instant with no EA connection. A refresh runs in the background from a single call, so a scheduled AI agent can start it and report the result. From any of the three ways, AI Power Tools generates a Power BI project with declared relationships, EA's own tables and the key columns hidden, a count measure on each entity table, a description on every business table, column and measure, and an Overview page. Published from Power BI Desktop, the model can answer questions in Microsoft 365 Copilot Chat, and a descriptions file you approve defines your business terms for it.

A scheduled AI agent starts a refresh, polls it and reports the result. Enterprise Architect must be running with the repository open on the machine where the refresh runs.

Prolaborate

Build, read and change Prolaborate dashboards. Read, filter and export relationship matrices. Trace dependencies and analyze the impact of a change with Prolaborate's Analyzer views. These skills use no server tools. They use browser control to drive the Prolaborate web interface, through the client's built-in browser or Chrome, in a session you have signed in to.

Getting started, support and fallback

Skill Role
ea-start-here Runs a short preflight at the start of a session, then routes to the right skill. Start here for any EA task
ea-diagnostic Produces a structured diagnostic report for support
ea-help Last resort. Finds the written procedure in Sparx's user guide for your EA version and carries it out in EA's user interface with computer use. Without computer use it hands you the steps
ea-com Drives EA from a script through its automation interface when no tool reaches the feature. The script runs through local code execution

Getting started and operations covers install, starting a session safely, reporting a problem and what to do when no tool reaches a feature.

AI Power Tools for Sparx EA · Release 3.6.4 Full reference for AI assistants (Markdown)