Agentic AI frameworks for industrial automation lifecycle management

Bosch Rexroth has integrated Agentic Artificial Intelligence into its ctrlX AUTOMATION platform. This platform-level integration establishes a central AI agent, known as iXi, capable of accessing automation data, engineering tools, and system functions to execute configuration and programming tasks throughout the automation lifecycle.

Transitioning from assistive to agentic AI in automation
Traditional AI assistants provide documentation explanations and generate code segments based on user prompts. Bosch Rexroth's agentic architecture advances this capability by enabling the AI to independently translate functional requirements into actionable system configurations. The AI agent acts as a centralized interface between the user and the automation environment. Rather than manually defining axes, configuring kinematics, and writing PLC programs step-by-step, engineers can provide the agent with a high-level system specification, such as a setup sketch and natural language description. The iXi agent then determines the necessary steps, defines the kinematics, and generates the corresponding IEC 61131-3 code directly within the engineering environment.

“Our aim is not only to develop the next co-pilot for PLC programming, but also for AI to understand automation and be able to actively work with it,” states Steffen Winkler, Head of Sales for the Automation & Electrification Solutions Business Unit at Bosch Rexroth.

System architecture and Model Context Protocol integration
To enable secure and standardized interaction between the AI agent and the automation infrastructure, Bosch Rexroth utilizes the Model Context Protocol (MCP). MCP establishes an encrypted and authenticated communication layer connecting Large Language Models (LLMs) with the ctrlX OS operating system, engineering tools like ctrlX PLC Engineering, and the ctrlX Data Layer. This architecture ensures the agent operates strictly within the permissions of the authenticated user, restricting unauthorized access to specific machine data and system functions.

Skill deployment and ecosystem scalability
The iXi agent operates on a "one agent, many skills" principle. Instead of deploying disparate AI models for individual tasks, the central agent queries specific, curated skill modules required for the immediate objective. If an operator requests a motion application, the agent retrieves the relevant motion control skill to accurately configure the axes within the ctrlX platform. This modular skill architecture scales dynamically; as new applications or third-party tools are added to the ctrlX ecosystem, they can inject their own specific skills, thereby expanding the agent's operational capabilities without requiring continuous retraining of the underlying LLM.

Inference flexibility and hardware development
The ctrlX AUTOMATION platform maintains an open architecture regarding AI model selection and inference location. Operators can integrate proprietary or third-party AI agents and LLMs based on their specific corporate infrastructure requirements. While cloud-based inference is currently supported, Bosch Rexroth is collaborating with AMD to develop a dedicated hardware platform for localized AI execution. This local inference capability will allow industrial operators to process sensitive engineering and machine telemetry data entirely on-site, mitigating the latency and data sovereignty concerns associated with cloud-dependent AI processing. The iXi agent is currently undergoing beta testing, with a broader market release scheduled for early 2027.

Additional Context
This section details technical specifications and competitive benchmarking not included in the original news release.

The integration of Agentic AI into industrial control platforms represents a shift from code-completion tools (such as GitHub Copilot) to autonomous configuration engines. Bosch Rexroth’s ctrlX AUTOMATION platform competes in this emerging space against Siemens' Industrial Edge ecosystem and Beckhoff's TwinCAT platform. While Siemens recently integrated generative AI into its TIA Portal via the Siemens Industrial Copilot (developed with Microsoft) to assist with SCL code generation, Bosch Rexroth’s utilization of the open-source Model Context Protocol (MCP) provides a distinct architectural approach. MCP allows the ctrlX platform to dynamically ingest context from third-party applications and engineering files without requiring rigid API integrations for every new tool. Furthermore, the collaboration with AMD for local inference addresses a primary barrier to AI adoption in manufacturing: data privacy. By processing LLMs locally on edge hardware, the ctrlX system can perform complex automated engineering tasks without exposing proprietary machine topologies or process data to external cloud networks, a requirement for many defense and high-security manufacturing sectors.

Edited by Aishwarya Mambet, Induportals Editor, with AI assistance.

www.boschrexroth.com