Changelog
See what we've been working on. New features and improvements every month.
Rebuilt from the ground up
A new engine and a new bridge to CODESYS-based IDEs and Beckhoff TwinCAT. It works inside the project you already have open, reads before it writes, and builds with the vendor's own compiler. Download the new bridge from the app — the old one no longer connects.
Your project, as a clear tree
The assistant now sees your project the way the IDE shows it: folders, function blocks, devices and libraries as one tree, so it finds the right place for new code and needs far fewer tokens to do it.
Pick up where you left off
Continue a chat where you left it, even after a reload. A step that was stopped or interrupted is reported as such instead of silently disappearing.
A steadier connection
The bridge reconnects on its own and the chat rides through short drops. When a change's outcome is uncertain, the assistant checks the project before doing anything else — it never sends the same change twice.
Clearer results, fewer tokens
Build results are grouped per item with the first errors first, and every tool answer is leaner, so long sessions stay fast and your allowance goes further.
A long list of fixes
From folder placement and naming to error messages and the status bar — many rough edges smoothed out across the app.
Pick your model — DeepSeek is the new default
New chats now run on DeepSeek V4.1 Flash, which handles most PLC work for a fraction of your monthly allowance. Every Claude model is still one click away in the picker, with its quota cost shown next to it. DeepSeek inference runs outside the EU — pick a Claude model for work you cannot send offshore.
Auto model selection removed
The chat used to pick a model for you. Now you choose, and the picker tells you what each model is for and what it costs against your plan.
More AI usage on every plan
Monthly AI allowances are up 70% on Pro, Power and the trial — same price. Power still gets 2.5x the usage of Pro.
Upgraded to Claude Sonnet 5
Code generation now runs on Anthropic's latest Sonnet model, with stronger multi-step reasoning on larger PLC projects.