Bosch Rexroth is integrating agentic AI at the platform level. AI is able to access the engineering tools, apps, documentation, as well as data and functions of ctrlX AUTOMATION and use them according to the context. This creates an architecture in which AI no longer works as an assistant alongside automation, but now works with it.
From assistant to agent
A classic AI assistant, such as the Ask iXi chatbot from Bosch Rexroth, is able to answer questions, explain documentation, or generate code. The capabilities of the new Agent iXi go far beyond this: The AI agent is able to translate tasks into individual steps and use the necessary functions of the automation environment. This difference becomes especially clear when it comes to coding: Through high-level languages, open APIs, and the ctrlX AUTOMATION SDK (Software Development Kit), ctrlX AUTOMATION already provides the basis for AI-assisted app and program development with coding assistants such as GitHub Copilot. With the new Model Context Protocol interfaces to the software tools ctrlX PLC Engineering and ctrlX I/O Engineering, Agent iXi enables much more: The AI agent generates IEC-61131-3 code directly in the engineering context and configures I/Os automatically. Assisted code generation thus becomes agentic engineering.
The AI agent thus becomes the link between the user and the automation environment. It can take on different tasks and access the required knowledge and functions for each of these tasks. These range from assisted component and topology selection, motion and drive setup, and I/O and HMI design, all the way to programming, system monitoring, and documentation.
The new possibilities can be illustrated with the example of a three-axis handling system. In classic engineering, axes have to be designed, kinematics configured, and the PLC program created. This requires many individual steps, such as configuring the coordinate system, motion limits, and axis weightings. In the new approach, Agent iXi receives a simple specification in the form of a drawing of the structure containing the key details, along with a description of the task in natural language. Based on this, it creates axes and kinematics and generates the PLC program for the desired motion.
"Users' work changes fundamentally: Instead of performing each technical step themselves, they increasingly only need to describe the desired result. The agent handles the necessary engineering steps and transforms them into a functioning automation solution, in a way that is always fully transparent to the user," explains Steffen Winkler, Senior Vice President Sales, Business Unit Automation & Electrification Solutions at Bosch Rexroth.
MCP provides the agent with secure access to automation
Bosch Rexroth uses the Model Context Protocol (MCP) to enable the agent to interact with the automation environment. The open standard connects large language models or AI agents with software and systems. Bosch Rexroth integrates this approach into ctrlX AUTOMATION and thus creates standardized access for AI to different areas of automation. For example, Agent iXi can access engineering tools, the operating system ctrlX OS, and the apps running on it. Within this process, the ctrlX Data Layer provides continuous and secure access to data and functions within the platform. Communication via MCP is encrypted and authenticated, and the agent also takes into account the permissions of the user that is logged in. This ensures precise control over which data and functions the agent, or the connected AI model, can access. This creates the technical connection that allows the agent to not only receive information but also work with the automation environment.
One agent – many skills
Skills provide the key context for automation. They give the agent the specific knowledge and rules it needs for a particular task. For example, if a motion application is to be created, the agent uses the corresponding curated skill and thus knows how to create axes and kinematics within ctrlX AUTOMATION.
Bosch Rexroth follows the principle of "one agent – many skills." Instead of using a separate agent for each task, the central agent accesses the precise capabilities it needs, depending on the task. This means it does not have to be provided with extensive context for every task.
The key scaling effect is that the architecture is designed to allow each app to bring its own skills. New or enhanced features within an app can thus simultaneously expand the agent’s capabilities. Third-party apps, too, can make their specific expertise available for use by AI in this way. This expands the range of available and validated AI capabilities with the open ecosystem of ctrlX AUTOMATION.
Open for agents, models, and inference
For Bosch Rexroth, AI-ready by design also means not restricting users to a particular AI world. In addition to its own agent, third-party agents can also be connected. The same principle applies to the underlying large language models. ctrlX AUTOMATION is not bound to a specific model. Companies can connect different models according to their requirements and existing AI infrastructure.
Bosch Rexroth is also committed to freedom of choice when it comes to the location of inference. In addition to cloud-based models, there are plans to enable the local execution of high-performance AI models in the future. Together with AMD, Bosch Rexroth is working on a hardware base on which AI can be run locally with ctrlX OS. This will enable sensitive engineering and machine data to be processed on site without necessarily having to leave the local environment for inference.
“For use, agentic AI must not end up in a new technological lock-in. Users should be able to decide whether and how they use AI, which agent and AI model they use, and where the inference takes place. We combine this freedom of choice with the automation context of ctrlX AUTOMATION. Thanks to the modular app concept, functions can be expanded flexibly at any time. And if you don't want to use artificial ntelligence, you can simply choose not to install the feature," says Steffen Winkler.
From engineering to additional phases of the machine life cycle
Engineering is already showing just how fundamentally agents can transform interaction with automation. However, Bosch Rexroth’s approach goes even further. ctrlX AUTOMATION is designed to make AI usable across different tasks and phases of the automation life cycle. Apps, data, and automation functions are interconnected in an AI-ready architecture. Bosch Rexroth is thus transforming AI from a support tool to a key player in automation – thereby creating the basis for agentic automation throughout the entire life cycle. Agent iXi is already being used by ctrlX AUTOMATION customers as part of a beta test. The test is expected to be expanded to include further customers over the next few months, and the AI agent is expected to be available in early 2027.
