SemaPLC is an MIT-licensed localhost web IDE and 16-tool MCP server from Midea AI that turns plain-language control specs into IEC 61131-3 Structured Text, compiles and runs it on OpenPLC Runtime v4 in Docker, forces inputs and traces variables to verify behavior, and renders a live ladder and process simulation. Everything runs locally apart from the model API you configure. Because it targets OpenPLC rather than proprietary vendor runtimes out of the box, treat it as a verification and prototyping harness before porting logic to plant hardware.
For controls engineers and OT teams: generating and behavior-testing IEC 61131-3 Structured Text against a local OpenPLC runtime before touching hardware.
Built in C# on .NET Framework 4.8 with a companion CLI, this open-source MCP server wraps Siemens TIA Portal Openness over stdio so an AI agent can open projects, manage hardware catalogs, blocks, UDTs, ProDiag and tag tables, import CSV or Excel lists, validate HMI targets, and compile software. It requires a Windows machine with a licensed Siemens TIA Portal V20 install and Openness user permissions, and the repo carries no open-source license file yet. Check licensing and run against offline project copies before letting an agent edit live station files.
For Siemens automation engineers: letting an MCP-capable coding agent inspect blocks, manage tag tables and run compiles in TIA Portal V20.
Atoms Not Bits reports that Anduril integrated its Lattice command software with existing air-defense radar and Origin Robotics' autonomous BLAZE interceptor in Latvia in less than four days to counter hostile drones and loitering munitions on NATO's eastern flank. The post shares no live intercept test metrics or contract value. For critical-infrastructure security teams, the notable detail is connecting legacy sensors to third-party autonomous interceptors in days rather than a multi-year integration program.
YC startup Liquid Energy says its factory-built NanoDC container unfolds solar panels, batteries, air cooling and Starlink on site so AI inference capacity can deploy in four weeks and turn on in 30 minutes without construction, water or a utility interconnect. The company plans its first live units this month and thousands starting in 2027, though per-box power output and unit costs are not disclosed. Even as an unaudited launch, off-grid modular compute shows how far buyers are looking to bypass multi-year transformer and interconnect queues.
