Munich, Germany – The momentum behind autonomous, multi-agent software engineering was on full display at the 41st IEEE/ACM International Conference on Automated Software Engineering (ASE 2026) in Munich, which gathered 933 registered researchers and industry leaders. A core highlight of the conference program was the 1st International Workshop on Agentic AI for Next-Generation Software Development (AgenticDev 2026), an event drawing around 100 participants dedicated to advancing the role of cooperative AI agents across the software lifecycle.
The full-day workshop addressed foundational, architectural, and operational challenges in modern software development across five dedicated technical sessions: Code Analysis, Agentic Coding, Testing, Repair, and Agentic Orchestration.
Uniting European Initiatives: The AGenSE Alliance Panel
MOSAICO took a central role in the workshop’s plenary programming through the AGenSE Alliance panel discussion. The panel brought together representatives from major Horizon Europe software engineering initiatives, including MOSAICO, ASTIR, AI4SWEng, and FAITH.
Representing MOSAICO on stage, Jean-Marie Mottu (IMT) and Tuong Vu (UY) outlined the project’s strategy for orchestrating communities of heterogeneous LLM agents. The discussion emphasized the necessity of European cross-project alignment on interoperability, human-in-the-loop governance, and verifiable safety guarantees to turn agentic development environments into reliable industrial tools.
Technical Contributions: Orchestration and Domain-Specific Modeling
Beyond panel discussions, MOSAICO demonstrated its core technical advancements with two peer-reviewed papers presented by Tuong Vu:
- “MOSAICO: A Multi-Agent Infrastructure for Collaborative LLM-Powered Software Engineering” (Agentic Orchestration Session)
This paper introduced the fundamental architecture of the MOSAICO platform, illustrating how multi-agent coordination layers, task delegation frameworks, and communication protocols enable specialized AI agents to jointly solve complex engineering workflows.
- “Towards Model-Driven Generation of Domain-Specific Modelling Agents for Local Language Models” (Agentic Coding Session)
Focusing on autonomy, efficiency, and data sovereignty, this work presented a model-driven approach to automatically synthesize domain-specific software modeling agents capable of running locally on smaller LLMs without relying on proprietary external APIs.
- “Defining and Deploying Agentic Swarms with the BESSER Low-Code Platform” (Agentic Coding Session)
Expanding the project’s reach, this collaborative research proudly acknowledged the support and vision of the MOSAICO project. The paper explored how the BESSER low-code environment can streamline the definition, deployment, and management of interacting agent swarms, aligning perfectly with MOSAICO’s goal of making complex agentic ecosystems more accessible.
With high attendance and vibrant technical exchange across all five paper sessions, AgenticDev 2026 underscored the rapid transition from simple coding assistants to structured, multi-agent development teams. MOSAICO continues to lead this shift, delivering open, scalable foundations for Europe’s next generation of software engineering.