FLAGSHIP · Bridge Automation · v0.1

Let your AI write PLC code.
You stay in control.

Connect any AI — Claude, ChatGPT, Gemini, local models — to your real PLC projects in CODESYS, Siemens TIA Portal and Mitsubishi. Every change is validated, backed up and approved by the engineer before it touches the machine.

How Bridge Automation works

A file-based bridge between the AI and the vendor IDE. Semi-auto by design: the AI proposes, the engineer approves.

SYNC TO AI → your live project exported for the AI to read AI proposes → a small, named change (one FB, one program, one GVL) Preflight → schema check + full backup + intent contract SYNC TO PLC → engineer approves inside the IDE — then it lands Snapshot → known-good state saved as readable source, any time
🔒

Safety contract

Automatic backup before every import. If the change doesn't land exactly as declared, the apply aborts — no silent overwrites, ever.

🎯

Partial by default

Import only the objects you changed. One function block, one program, one variable list — never the whole project by accident.

📸

Golden snapshots

One click freezes a known-good state: a restorable backup plus per-object plain-text source you can read and diff without opening the IDE.

🤖

Any AI, no lock-in

A simple file protocol — works with any AI that can read and write files. No cloud dependency, no subscription to someone else's model.

PlatformStatus
CODESYS V3 (incl. Delta AX, Schneider, WAGO, ABB) ✅ Full loop — proven
Siemens TIA Portal (SCL / LAD) ✅ Proven — SCL both ways + LAD (FC), auto-apply via Openness
Mitsubishi GX Works3 ✅ Proven — ST / FB / ladder · one manual import step
Factory I/O scene analysis 🔶 Preview
Factory-A

Factory-A — verified AI coding

Let any AI ship code that actually works. Every change passes a hard-coded gate before it’s declared done. Your project’s memory is portable across sessions and AI vendors.

🎯

Deterministic gate

No hallucinated names → tests 100% → replay-deterministic → tamper-evident receipt. The verifier is a subprocess, not another LLM. If it doesn’t pass, it doesn’t ship.

🧠

Append-only memory

YAML knowledge nodes with reproducibility contract. Session A saves; Session B (any vendor) reads. What you learn stays learned — never overwritten.

🔌

Any AI client

Bundled MCP server — Claude Code, Claude Desktop, Cursor, Codex CLI, VS Code Copilot. Six real tools, all execute; no vendor lock-in, no cloud dependency.

🧩

Multi-module wiring gate

Cross-module contract shape match + import existence check + topological build order. Catches the silent bug where module A renames a field and module B breaks.

MIT license · Works with any MCP-capable AI client · Python 3.10+

Download Bridge Automation — 30-day free trial

Full features for 30 days. No account, no card — install and run. Windows.

⬇ Download Bridge Automation  Windows · ~1 MB

Trial runs 30 days from first launch; after that, SYNC TO PLC / SYNC TO AI need a full license.

1

Install Python 3.11

Required — the ladder/digest engine is a CPython-3.11 module. Tick "Add python.exe to PATH". Get it at python.org 3.11.

2

Unzip & INSTALL.bat

Double-click INSTALL.bat → a Bridge icon appears on your Desktop.

3

Drive it with your AI

Add your vendor, click SYNC TO AI, and let your AI (Cursor / Claude / Codex / VS Code) write the PLC code. You approve every apply.

Products

Tools built in a real factory-automation practice — shipped when they survive real machines.

Bridge Automation available now

The AI ↔ PLC bridge described above. One-time license, activate on your machine, offline-friendly for air-gapped factory PCs. Tutorials included. Download the 30-day trial →

LXR Platform in development

The namesake: a deterministic platform for isometric mechanism design — author a machine as a single-line L/R skeleton and generate kinematics, step logic and PLC code from one source. Specification-first, machine-verified.

Factory-A available now

A verification framework for AI-driven software work: deterministic gate, append-only knowledge graph, bundled MCP server. Free and open source. See how it works →

Contact

Early access, questions, or a machine that needs taming:

📧 incott.th@gmail.com

LXR