Webinar Transcript Summary
Sept 2026 | Stéphane Lacrampe (Obeo) | EN
Artificial intelligence offers new opportunities for Model-Based Systems Engineering (MBSE), particularly in accessing, analyzing, and maintaining complex system architectures. However, its adoption raises important questions about reliability, traceability, and engineering responsibility.
Obeo's new AI for Capella solution addresses these challenges by enabling AI agents to interact directly with Capella models while preserving their consistency and integrity. Rather than replacing systems engineers, the objective is to simplify modeling activities, accelerate analysis, and support better-informed engineering decisions.
Capella models provide a particularly valuable foundation for AI because they contain structured engineering knowledge based on the Arcadia methodology.
Unlike unstructured documents, these models explicitly describe capabilities, functions, components, interfaces, dependencies, and relationships across engineering viewpoints.
This structured information enables AI agents to interpret engineering questions, navigate complex architectures, identify dependencies, and detect potential inconsistencies.
The main objective is to reduce time spent on repetitive modeling operations and information retrieval, allowing engineers to focus on architectural decisions and system design.
AI for Capella adopts a Bring Your Own AI approach, allowing organizations to use their preferred AI environments, including ChatGPT, Claude, or locally hosted language models.
The integration relies on the Model Context Protocol (MCP), an open standard connecting AI agents to external tools.
An MCP server installed within Capella exposes controlled modeling capabilities to AI agents. These agents can query, create, or modify model elements through Capella's native mechanisms rather than manipulating model data directly.
Access permissions, including read-only and editing profiles, ensure that engineers retain control over authorized operations.
The architecture also enables AI agents to combine Capella information with external engineering repositories, including requirements management, PLM, and documentation systems, when suitable connections are available.
The webinar illustrated several practical applications using Capella models.
Model Creation and Onboarding: An AI agent guided a new user through Arcadia concepts and automatically created an optical mouse architecture, including components, functions, exchanges, diagrams, and functional chains.
Architecture Exploration: Using an In-Flight Entertainment System model, the agent generated accessible system summaries, explained component responsibilities, traced capabilities to physical equipment, and identified incomplete architectural relationships.
Impact Analysis and Consistency: The agent evaluated potential consequences of requirement changes, aircraft platform modifications, and equipment failures. It also detected inconsistencies between system and logical architecture states, including missing safety-related behavior.
Architecture Evolution: A new Wi-Fi capability was introduced by identifying reusable components, creating additional actors and functions, and updating architectural diagrams.
Requirements and Documentation: The AI generated preliminary design requirements and review documentation. It also demonstrated the possibility of producing M2Doc templates for repeatable, model-based document generation.
Automotive System Design: Starting from an Excel spreadsheet containing supplier components, requirements, and technical constraints, the agent generated an automatic emergency braking architecture. It then evaluated mass, cost, power consumption, and latency, identifying a latency exceeding the project target by 26 milliseconds.
These demonstrations highlighted the potential of AI to accelerate modeling and analysis while also revealing the importance of precise prompts and engineering validation.
AI for Capella supports several Capella versions and follows a regular development cycle, with releases approximately every eight weeks.
Upcoming priorities include improved reliability, simplified installation, enhanced documentation, broader AI compatibility, and additional modeling automation capabilities.
Future developments will also extend AI to Publication for Capella, enabling stakeholders to explore published architecture models through natural-language queries without installing Capella.