Equipment experts, not coders
You know ladder logic inside out, but Python, MQTT and GUI frameworks are a different skill set that takes months to learn.
AI for OT · AI programming for OT engineers
OT engineers can ship working software without becoming developers
The people who know the equipment best are on the OT side: PLC, maintenance and automation engineers. Falcon turns PLC ↔ HMI traffic into JSON-RPC data an AI can understand. Describe what you need in your own terms and an AI agent writes the code, then tests it against replayed data until it passes.
Where OT engineers get stuck
Waiting on IT
A monitoring screen or a report can sit in the IT queue for months, and the requirements often get lost along the way.
No safe place to test
A bug can disrupt production, so testing waits for shutdowns and maintenance windows that come a few times a year.
Falcon + AI Agent
What Falcon and an AI agent change
An AI needs data it can understand and a test environment where mistakes are harmless. Falcon provides both.
Speak shop-floor language
Just say “D100 is press pressure in 0.1 bar.” Tags arrive as JSON-RPC, so the AI can map your words to real data.
Log replay is a safe sandbox
Development and testing run on recorded field data. If the AI gets it wrong, only your laptop notices. The line never does.
The AI tests until it passes
The agent runs the code, compares it with the replay and fixes its own errors. You receive a verified result.
You own the code
The output is standard Python and PySide6 you can read and change, with no lock-in to a closed platform.
A real example
Use the words you already use on the floor. No programming vocabulary required.
Record field data
Capture a real production log with Falcon.
Describe it your way
Tell the AI agent what the tags mean and what you want to see.
AI writes and tests on replay
The agent codes, replays and fixes until everything passes.
You approve go-live
Check the result on the replay, then point it at the site MQTT broker.
ai-agent · ot-request
> D100 is press pressure in 0.1 bar; M20 is main motor running. If pressure stays above 120 bar for 3 seconds, flash a red lamp on screen and log the time and pressure to CSV.
● Found D100 and M20 in the replay log; D100 ranges 0–1350 (= 0–135.0 bar)
● Wrote alarm logic, red lamp widget and CSV logging (PySide6 + paho-mqtt)
● Ran log.replay on shift-a: detected 2 overpressure events
↻ Event 2 lasted only 2.4 s but raised an alarm: fixed the timer and retested
✓ Done: 1 alarm, 1 CSV row, matching the replayed data
Use cases
What OT engineers can ask the AI to build
Work that used to need IT or a contractor, done by the people who know the process.
Remote HMI screens
Bring key values from the local HMI to the office or a manager’s desk.
Data logging and reports
Record tag values on a schedule and export daily or shift reports to CSV or Excel.
Alarm notifications
Notify people when values leave limits, run too long or change state unexpectedly.
Availability and OEE
Calculate availability and OEE from running, stop and count tags.
Tag inventories
Build a list of the tags the HMI really uses, with value ranges, from recorded traffic.
Before/after PLC checks
Replay the same data to compare behavior before and after a PLC program change.
Safety first: the AI never touches the PLC
- Development and testing use replayed data only, with no connection to the line
- Generated code reaches tags only through Falcon JSON-RPC methods, never the PLC directly
- Writing tags needs separate authorization, enforced with MQTT credentials and topic ACLs
- OT engineers decide what goes live; the AI only writes and tests
FAQ
I don’t know Python. Can I still build tools with AI?
Yes. Describe the requirement in your own terms, such as what each tag means and what should happen under which conditions. The AI agent writes and tests the code; you confirm the result on screen.
Which AI agents work with Falcon?
Any AI coding agent that can write and run Python, such as Claude Code. Falcon exposes standard JSON-RPC over MQTT and is not tied to a specific AI service.
Can AI-generated code change the PLC directly?
No. Development uses recorded log replay only. In production, code can reach tags only through Falcon JSON-RPC methods, and writes need separate authorization.
Can I modify the generated code myself?
Yes. It is standard Python and PySide6 code you can edit, hand to a colleague or ask the AI to adjust.
How is this different from traditional SCADA or HMI software?
Traditional packages have fixed features and customization goes through the vendor or an integrator. Falcon with AI lets OT engineers build purpose-made tools quickly and validate them fully on replayed data.
Get your line speaking MQTT today
Tell us your PLC / HMI models and site requirements. Our sales team replies with an integration assessment and quote within one business day.