Features

1. What Obeo AI for Capella Does

Obeo AI for Capella connects compatible AI assistants to a running Capella environment so engineers can explore, understand, analyze, modify, document, and verify architecture models.

The recommended workflow is:

Explore -> Understand -> Analyze -> Decide -> Modify -> Verify

2. What You Can Do

Capability What you can do Example request

Discover

Find available Capella environments, projects, models, and diagrams.

“Show me the projects and diagrams available in this workspace.”

Understand

Explore model structure, diagram contents, components, interfaces, dependencies, and existing documentation.

“Explain the architecture around this component and its interfaces.”

Analyze

Assess change impacts, affected interactions, risks, and possible architecture improvements.

“What would be affected if this component responsibility changed?”

Modify

Create projects, representations, model elements, relationships, and diagram content through controlled operations.

“Add a logical component to this diagram and connect it to the existing system.”

Document

Read and update model or diagram documentation and maintain realization links.

“Document this component’s responsibility and link it to the realized element.”

Verify

Refresh, arrange, inspect, and export diagrams so changes can be reviewed.

“Refresh and arrange this diagram, then export it as a PNG for review.”

3. Tool Reference

Most users can describe their goal directly to the AI assistant without choosing a tool themselves. This compact reference is intended for integrators and advanced users. The minimum profile is the least permissive server profile that publishes the tool. See the MCP reference for profile configuration, metadata, and detailed execution behavior.

Evolution note: The proposed tool surface and its classification concepts, including purpose, layer, intent, mutating behavior, profile, and batch execution, are expected to evolve in upcoming versions.

Tool Purpose Minimum profile

capella_get_agent_guidance

Return the embedded operating contract and Arcadia context for clients that cannot expose MCP resources to the model.

read-only

sirius_list_runtimes

List discoverable running automation environments.

read-only

sirius_list_projects

List projects in a selected workspace.

read-only

sirius_get_session_details

Return model sessions, semantic roots, representations, and collaborative state.

read-only

capella_create_project

Create a Capella project in the selected workspace.

edit

sirius_list_representation_descriptions

List representation definitions compatible with a model context.

read-only

sirius_list_representation_candidates

List semantic owners for a selected representation definition.

read-only

sirius_list_session_representation_descriptions

Discover representation definitions across a model session.

read-only

sirius_list_session_representation_candidates

Discover compatible semantic owners across a model session.

read-only

sirius_manage_representations

Create, open, or close representations.

edit

capella_inspect_model

Summarize model structure, architecture packages, and validation context.

read-only

sirius_describe_diagram

Describe diagram views and available creation and selection palette operations.

read-only

sirius_list_palette_tool_candidates

List semantic elements accepted by a discovered palette selection tool.

read-only

sirius_get_active_selection

Read the active diagram or workspace selection.

read-only

sirius_set_active_selection

Set the active diagram or workspace selection.

edit

sirius_apply_palette_tools

Execute discovered palette operations; prefer this route when newly created content must be visible in a diagram.

edit

capella_list_element_creation_actions

List direct-child semantic creation actions for explicit semantic-only intent or diagram fallback.

read-only

capella_execute_element_creation_action

Execute one semantic creation action without creating or guaranteeing a diagram view.

edit

capella_analyze_impact

Return semantic references and affected representations for an element.

read-only

sirius_manage_diagrams

Arrange, refresh, export, or rename content in diagrams.

edit

sirius_get_diagram_description

Read documentation stored on a diagram.

read-only

sirius_set_diagram_description

Replace documentation stored on a diagram.

edit

capella_arrange_diagram_optimized

Arrange an open Capella diagram with optimized ELK layout.

edit

capella_apply_model_edits

Attach realization links through ordered Capella edit steps.

edit

capella_get_element_description

Read documentation from the selected Capella element.

read-only

capella_set_element_description

Update documentation on the selected Capella element.

edit

capella_resolve_part_abstract_type

Resolve the typed component behind a Capella Part.

read-only

sirius_execute_aql

Query one specific missing fact that dedicated inspection does not expose, or run explicitly requested read-only AQL.

read-only

emf_describe_model

Inspect EMF types, filtered properties and typed references; navigate session roots and direct containment.

read-only

emf_apply_edits

Create Ecore metamodel elements or edit writable features of existing Capella elements.

edit

Prefer Capella or Sirius tools for Capella model changes; use direct feature edits only when those tools do not express the requested change.

4. Legacy Tools

The following older tools remain available for compatibility. Legacy tools are available only under edit; they are not guaranteed to leave data unchanged. New workflows should use the recommended alternatives.

Legacy tool Recommended alternative

capella_describe_logical_component

Use capella_inspect_model, sirius_describe_diagram, or emf_describe_model; use AQL only for a named missing fact.

capella_list_allocated_logical_functions

Use emf_describe_model, then focused AQL only if the allocation fact remains unavailable.

capella_list_connected_logical_components

Use sirius_describe_diagram or emf_describe_model, then focused AQL only for a missing connection fact.

capella_list_component_interfaces

Use sirius_describe_diagram or emf_describe_model, then focused AQL only for a missing interface fact.

capella_trace_functional_dependencies

Use emf_describe_model, then focused AQL only if the dependency fact remains unavailable.

capella_analyze_logical_impact

Use capella_analyze_impact together with model and diagram inspection.

capella_propagate_logical_impact

Use capella_analyze_impact, then let the assistant synthesize the propagated impact.

capella_summarize_impact_table

Use capella_analyze_impact with generic inspection, then let the assistant synthesize the table.

capella_analyze_impacted_interactions

Use capella_analyze_impact with generic inspection, then let the assistant analyze interactions.

capella_suggest_architecture_improvements

Use capella_analyze_impact with generic inspection, then let the assistant propose improvements.

capella_summarize_engineering_decision

Use capella_analyze_impact with generic inspection, then let the assistant summarize the evidence.

5. Multi-Step Operations

When an operation contains several steps, execution stops at the first error. For operations that change the model, completed changes remain applied and are not automatically rolled back. See the MCP reference for detailed retry and execution guidance.

6. Deployment and Compatibility

The following client and transport combinations have been tried. This list is not an exhaustive compatibility guarantee.

Transport Clients tried

HTTP

Codex CLI; Gemini CLI (Enterprise/API); Qodo IDE plugin; Kiro IDE

stdio

ChatGPT desktop app / Codex CLI; Claude Code / Claude Desktop; Antigravity IDE; VS Code Chat on Windows

For client setup examples, see LLM Client Configuration.

The available trial records did not distinguish between the listed Codex or Claude surfaces for stdio. Check the exact surface before treating either stdio entry as a surface-specific compatibility guarantee.

The default transport is stdio. Local HTTP binds to 127.0.0.1 by default. Remote plain HTTP is rejected; remote HTTPS requires a bearer token. The server also supports allowed-origin and allowed-host restrictions.

The MCP and CLI tools require Java 21, included with standard Capella 7.1.0. Installation targets Capella 7.1.0 and also supports 7.0.1 and 6.1.0 on Windows, Linux, and macOS. Older Capella versions and SMW distributions without a bundled Java 21 need a separate Java 21 for these tools; see the Java setup instructions.

Local deployment does not guarantee that model data remains on the local machine: the configured AI client or model provider determines what context is sent outside the environment.

7. Demonstration Scenarios

  • The Airport Turnaround acceptance scenario exercises representation and diagram operations together with Capella inspection and analysis.

  • The IFE USB integration scenario covers component understanding, impact analysis, architecture evolution, model editing, verification, and option comparison.

  • Diagram scenarios demonstrate arrangement and PNG export as reusable evidence.