Reads your TIA project as a graph
Every block, caller, cross reference and your custom libraries.
PLC Coding Agent
Your PLC team.
Just faster.
An AI agent that understands your TIA Portal project, proposes verified SCL code and backs every change with compile and test evidence. Your programmer stays in control.
No commitment. 30 minutes. Directly with a founder.
Floorstack Agent | Project: AssemblyLine_04
You: Why does the conveyor not restart after an emergency stop?
Agent: FB_Foerderer expects a rising edge on #Quittierung (network 3). FC_Bedienpanel stopped setting it after the change from May 12. Proposal below, affected callers: 2.
IF #Start AND #Betriebsbereit THEN
- #Motor := TRUE;
+ #Anlauf(IN := #Start, PT := T#2S);
+ #Motor := #Anlauf.Q;
END_IF;
✓ TIA compile✓ PLCSIM test 4/4✓ Lint 0 errors
ApproveReject
Schematic view of the review interface
Trusted by
The product
The agent knows your project better than any new hire.
Not a chatbot guessing SCL snippets. An agent that works on your actual project state and backs every proposal with evidence.
Answers questions with citations
Like a colleague who knows every network by heart. "What sets this bit? Who calls this FB?"
Proposes changes as a diff
Reviewable like a clean change request, including an impact analysis of affected blocks.
Proves instead of claiming
Every change is compiled in TIA Portal and checked against behavior tests in PLCSIM Advanced before a human sees it. Every result comes from an actual test run and is recorded in the evidence protocol.
Platforms
Our core: Siemens TIA Portal (SCL)
In progress: Beckhoff TwinCAT and CODESYS. Support is not tied to one TIA version, pilot partners set the order.
The ground rule
The agent cannot do anything your programmer has not approved.
Day to day
Three situations you know.
Series machine, new variant
The variant change
A customer orders the machine with an additional station. Today: days of adapting blocks, HMI and docs across the whole project.
With the agent: It knows every variant location, proposes the complete change set, your programmer reviews and approves.
Legacy line, the author is gone
The inherited project
A project from eight years ago, the author has left, the docs are thin. Today: days of clicking through networks.
With the agent: Ask instead of search. "What sets this bit? Who calls this FB?" Answers with citations in seconds.
Before commissioning
The review under deadline
The line ships to the customer next week. Today: reviews and standard checks get skipped, defects travel to the site.
With the agent: Impact analysis, lints and PLCSIM tests run in the office beforehand. Fewer surprises on site.
Your concrete use case is defined together at the start of the pilot.
Safety
Safety is architecture here, not a promise.
Human gate
Every live change requires your programmer's approval. Refusing is the default.
Checkpoint and auto rollback
A restore point before every import, automatic revert on compile errors.
Evidence over claims
Lint, compile, test and review results come from actual runs.
Safety and F blocks are off limits
The agent never touches safety rated logic.
Confidentiality is standard
NDA before the first demo if you wish. Your project data never goes to third parties.
The human stays responsible
The agent delivers evidence, not excuses.
Evidence protocol | Change #217
Add 2s soft-start ramp to conveyor
Change in FB_Foerderer | callers checked: FC_Transport, OB1 | no access to safety blocks
✓ Lint passed✓ Compile clean✓ 4 behavior tests green
Checkpoint setRollback ready
Schematic view. Every change produces a protocol like this from actual test runs.
Exactly this architecture separates an industrial agent from a chatbot guessing SCL snippets.
The offer
The pilot: 3 months from kickoff to result.
- 1
Month 1
Setup and daily work
Your TIA project as an export or test copy, anonymized if you prefer. The agent learns your project structure and your libraries, the use case is defined together, your PLC team starts working day to day.
- 2
Month 2
Iteration
Weekly feedback, adjustments within days, influence on the roadmap.
- 3
Month 3
Result
Review with evidence protocols, joint rollout decision.
Kickoff on site
We start at your plant: define the use case, hand over the project, meet the team.
Weekly update meeting
One fixed meeting per week: progress, open items, next steps. It takes no more than 1 to 2 hours of your team’s time per week.
You set the priorities
You decide which systems and features come first. Your requirements steer the pilot roadmap.
Adoption comes first
Success means your PLC team uses the agent day to day. We support the rollout until that is the case.
Who is behind this
Three founders, TU Munich and Stanford. Reachable when it matters.
In the pilot you work directly with the founders, not with a ticket system.
TU Munich | research at Stanford | previously Langdock
TU Munich | Cambridge | MIT | TDK incubator
TU Munich | ICML publications | Fraunhofer, IBM
Next step
30 minutes. Live demo with the CPO.
You bring a question from your daily work, Felix shows the agent live in TIA Portal. That is enough to define the use case for a pilot. We handle everything else.
Felix Hadasch
CPO & Co-Founder
Runs your demo personally and answers technical questions directly.
Prefer email?
felix@floorstack.aiBook a slot directly
The booking calendar loads from Google only after your click. No data is transferred to Google before that.