Software automation
Stateful logic over HTTP, files, and structured data. State survives a restart; every value you record lands in the historian.
Software, devices & AI
Connect your systems.See how they behave.
A stateful runtime with 2,700+ built-in functions: HTTP, WebSockets, files, JSON, CSV, databases, MQTT, crypto, shell. Program it in Structured Text, drive it from the browser IDE or the API, and let an AI agent work through the same tool surface you do.
Runs on a laptop, a server, or a Raspberry Pi. No AI required to run it.
Beyond the factory
Automation is bigger than PLCs.
Start with the behavior you want, not the category you belong to. ControlForge puts programming, durable state, and operating evidence in one environment, and it already speaks to the things you want to connect.
Connected projects
Test equipment, interactive installations, creative rigs. Microcontrollers over serial, lighting over DMX, cameras through the vision functions.
AI-assisted development
An agent introspects the runtime's real function list through MCP, drafts Structured Text, and gets it validated before it runs. Claude, OpenAI, and Ollama are wired into the IDE.
Example workflow
Turn a useful automation into a better toolkit.
An application receives structured data, applies stateful rules, and produces a result. Then one day it produces the wrong one. That failure is the start of a better application and a better process.
- 01 / BUILD
Define the behavior.
A small input-to-output path with explicit states. Use a real integration and a local test source so you can run it a hundred times.
- 02 / OBSERVE
Inspect the failure.
Watch state and diagnostics live, then trend the history. Tell a bad input from a dropped connection from a logic error, with evidence.
- 03 / IMPROVE
Test your change.
Have AI explain or revise if that helps. Then check the result against tests you control, not the model's confidence.
- 04 / REUSE
Keep what worked.
A reusable block, the test data, and a better diagnostic view. Start the next automation from there.
The compounding gain is yours: a better application now, and better tools and understanding for what comes next.
The programming model
Structured Text, without the mystery.
One language for all of it. Structured Text is small, explicit, and readable by the person who has to debug it at 2 a.m. Work in the browser IDE or through the API. Start from a working project and change its behavior with its state in front of you.
The role of AI
Assistance, not a prerequisite.
Use AI to draft, explain, and investigate, or let an agent work through the MCP tool surface on its own. The runtime treats its output like anyone's: validated before it runs, recorded when it does. A validated program is one the runtime can run. Whether it does the right thing is what your tests are for.
Proof, not promises
Built, filmed, written up.
Two complete project walkthroughs with bill of materials and code, and two videos of the runtime doing the work.
Project walkthrough
A robot dog on AI-generated ST.
Megabite: OLED eyes, servo ears, head tilt, driven by AI-written Structured Text on a Parallax P2.
See itProject walkthrough
A laser cutter over G-code.
From raw HTTP_GET calls to the full G-code function library, driving an xTool P2S from Structured Text.
See itVideo
AI writes the Structured Text.
Claude, OpenAI, and Ollama generating IEC 61131-3 ST from a plain-language description, inside the IDE.
See itVideo
Node-RED dashboards and flows.
The runtime manages Node-RED as a subprocess for operator dashboards and external connections.
See itBuild something you can build on
What would a better starting point unlock?
Tell us about the software, devices, or workflow you want to connect. ControlForge is in private access.