multi-plc-agent runs one MCP server that lets an AI agent write ladder logic in nine PLC tools, including GX Works2, TIA Portal V20, CX-Programmer, Sysmac Studio and CODESYS. A tool is checked off only after a real program built with 0 errors and read back matched. Tools that act on a live PLC or delete project content are not exposed, so the agent writes and compiles offline.
For controls and automation engineers: testing AI-written ladder logic offline.
ethercat-mcp is an MCP server that lets an AI assistant scan EtherCAT slaves, step the state machine, read and write CoE objects and PDO data, and diagnose bus health. It ships with a five-slave virtual bus of Beckhoff and Inovance devices, so it can be tested without hardware. It also exposes CiA 402 servo commands like enable, jog and positioning, so keep it on the simulator and away from a live line.
For controls engineers: who want to test agent tooling against a simulated EtherCAT bus.
AgentCAD is a CAD runtime an AI agent calls over MCP, a CLI or Python to create and edit DXF drawings and parametric FreeCAD parts. Each change is checkpointed and reversible, and a separate validator re-reads the file to flag geometry and dimension problems. It is a day old, so watch it before any drafting team relies on it.
For drafters and design engineers: curious about agent-driven drawing edits.
Otto Group will deploy ten AI-controlled dual-arm robots from Sereact to sort returns at its logistics center in Poland, with 15 more planned based on first results for a total of 25. Sereact runs and maintains them as a service, and staff move to supervision, exceptions and quality checks. Results-gated rollouts are how robot deals are getting signed.
