Autonomy CoPilot
AI that was taught industrial automation on purpose.
General AI models miss proprietary industrial code. Built by engineers for engineers, CoPilot delivers a seamless experience across Autonomy Edge in your browser and the OpenPLC Editor on your machine.
- 3
- Languages written: ST, Python and C++, with Ladder coming
- Whole project
- In context before it answers, so the answer fits yours
- Community
- Included from this tier up
It does not suggest. It builds.
Not a chat box bolted onto an editor: CoPilot holds the same model of your project the editor does. One request, and it writes the function block, declares the interface, allocates the variables and wires the instance into main.
Autocomplete as you type
It intelligently predicts and completes the code you are writing in real time, drawing directly from the context of your industrial logic to deliver instant, accurate suggestions as you type.
A chat, and you approve what lands
Ask for what you need in a conversation. CoPilot implements it, and the result waits for you to review it and approve it before it goes into the project.
| # | Name | Class | Type | Location | |||
|---|---|---|---|---|---|---|---|
| 0 | BlinkTimer | Local | TON | ||||
| 1 | PurpleTimer | Local | TON | ||||
| 2 | GreenLED | Local | BOOL | %QX0.0 | |||
| 3 | PurpleLED | Local | BOOL | %QX0.1 |
Ask about your project
The AI can create POUs, add variables, explain code, and edit your project.
Ask about your PLC code...
CoPilot knows your project
Six things a general-purpose model cannot do for you.
Whole project context
It reads every POU, variable, type and library before it answers.
Writes the POUs
Programs, Functions and Function Blocks, in ST, Python or C++.
Allocates the variables
Right scope, right type, an address that does not collide. Structures, enumerations and arrays too.
Explains what you inherited
Ask what a block does. The answer comes from your code, not from a manual.
Knows the libraries
TON, TOF, CTU, PID, RAMP, MQTT and TCP/IP, OSCAT and PLCopen MotionControl.
Determinism untouched
AI writes the logic. The runtime still executes it in deterministic scan cycles.
Where this is going
How CoPilot is being taught the job
We are building the data corpus that would make an AI genuinely good at IEC 61131-3.
Ladder is next
Ladder is what most of this industry actually writes, and it is the next language CoPilot will write. Same whole-project context, same tools, applied to the language the plant floor runs on.
The corpus nobody had
Off-the-shelf models learned from decades of public IT code. Industrial control never had that, so a general model returns ST that looks convincing, does not compile, and ignores the scan cycle it has to run in.
Somewhere for it to live
A corpus needs a place where industrial code is written, shared, compiled and corrected in the open. The open stack and the Community tier are that place, and every project made in them adds to the record.
A loop a closed vendor cannot buy
Real IEC 61131-3 projects, the errors the compiler raised and the changes that fixed them. That is what we train and fine-tune on, and it compounds with every engineer who joins.
Trained on the work, not the internet
CoPilot is tuned on industrial control specifically, rather than a general model asked politely to pretend it has seen a function block before.
Built by people who ship control systems
The team behind OpenPLC has been building PLC tooling since 2014. We are training a model on work we do, not on an industry we read about.