Vector CANoe integrates AI agents for automated testing workflows

Vector has expanded its CANoe development and testing environment with new artificial intelligence (AI) and Model Context Protocol (MCP) capabilities. Available in Version 20 SP2, the integrated MCP Server and CANoe AI Package allow users to execute complex tasks—from generating CAPL tests to analyzing communication flows—using natural language prompts. The specialized AI agents automatically derive the necessary steps and execute complete workflows, reducing tasks that previously took hours or days to just minutes.

Users can determine the level of autonomy for the agents and monitor every step live within the development environment. The CANoe AI Package relies on an open ecosystem where users supply their own underlying language model (such as Claude or the LLM behind GitHub Copilot), while Vector provides the open AI layer consisting of agents, skills, and MCP tools. Through Vector-RAG (retrieval-augmented generation), these agents draw from a verified knowledge base of Vector documentation, ensuring that outputs are grounded in expert knowledge rather than model assumptions. The free package enables both new and experienced users to seamlessly orchestrate customized, automated testing processes.

Additional Context
This section provides technological and market background not explicitly detailed in the original release.

The Model Context Protocol (MCP) has recently emerged as a standardized, transport-agnostic framework designed to define how AI agents securely obtain, exchange, and act upon contextual data. In automotive software engineering, network simulation and validation have historically required extensive manual scripting in specialized languages like CAPL (CAN Access Programming Language). By introducing an MCP server directly into CANoe, Vector bridges the gap between general-purpose Large Language Models (LLMs) and highly specialized automotive simulation environments. This "agentic" approach allows organizations to leverage their enterprise-approved LLMs to safely read proprietary test configurations, compile code, and run interactive simulations, significantly accelerating software-defined vehicle development without exposing internal IP to public networks.

Edited by Lekshman Ramdas, Induportals editor – adapted by AI.

www.vector.com