Vector Informatik has added AI agents and Model Context Protocol (MCP) support to its CANoe development and testing environment, available with CANoe 20 SP2. Users can describe development, analysis, and testing tasks in natural language, and AI agents determine and execute the required steps, such as converting a software requirement into a CAPL test, running it, analyzing failures, fixing the code, and rerunning it. The language model is user-supplied (e.g., GitHub Copilot or Claude), while Vector supplies agents, skills, MCP tools, and a Vector-RAG mechanism grounding responses in Vector's documentation. Automation autonomy is configurable, with real-time monitoring and human validation of results. The CANoe AI Package is a free download compatible with CANoe 20 SP2 and later.
Questions this post answers
What new AI capabilities does CANoe 20 SP2 add?
CANoe 20 SP2 introduces the CANoe AI Package with an integrated MCP Server, letting users describe development, analysis, and testing tasks in natural language. AI agents can convert a software requirement into a CAPL test, execute it, analyze failures, modify the code, and rerun it, presenting results through synchronized CANoe windows for validation. daily.dev surfaces updates like this for teams automating embedded test workflows with AI agents.
Which language models work with Vector's CANoe AI Package?
The CANoe AI Package requires a user-provided language model, such as GitHub Copilot or Claude, rather than shipping its own. Vector supplies the surrounding AI layer, including agents, skills, and MCP tools, plus a Vector-RAG mechanism that grounds outputs in Vector's own documentation instead of relying solely on the model's general knowledge. Engineers weighing AI model choices for engineering tools can track these integrations on daily.dev.
How much does Vector's CANoe AI Package cost and what version is required?
The CANoe AI Package is a free download from Vector, compatible with CANoe 20 SP2 and later versions. It still requires users to supply their own language model, such as GitHub Copilot or Claude, since Vector provides only the AI agent layer, MCP tools, and Vector-RAG knowledge grounding rather than the underlying model itself. daily.dev helps engineers keep up with tooling updates before rolling out new CANoe versions.