AI in ScaliRo · from version 1.32
Ask your layout.
An assistant right inside the Editor, an MCP interface for your own AI, and documentation that answers back. The AI takes the clicking off your hands. How many robots you need is still worked out by the simulation.
Example illustration
What is AI in ScaliRo?
Three ways in. One project.
AI in ScaliRo is not an add-on product. It is three ways into the same project data: the assistant in the Editor, the MCP interface for your own AI, and the docs AI. Use exactly the one you need.
Assistant in the Editor
- Knows your selection
- Checks patterns across hundreds of stations
- Changes nothing until you approve
For example “Do all transfer stations follow the same pattern as these three?”Learn more
MCP interface
- The same tools as the assistant
- For Claude Code, Codex & co.
- A key that works only for MCP
For example “Simulate one 8 h shift at 10x, then pause.”Learn more
Docs AI
- Answers from the documentation
- Names its sources
- Says when something is not in there
For example “Which DXF files can I import?”Learn more
Assistant in the Editor · from 1.32 · early access
Your assistant for
the clicking.
It works in the project you have open. You describe what you need in one sentence. It proposes, you approve.
You mark three, it finds the rest. No typing IDs.
Proposal, approval, one undo step.
- Transfer 05 · X offset
257 mm250 mm - Transfer 10 · X offset
257 mm250 mm - Transfer 13 · X offset
257 mm250 mm
Hundreds of properties at once, one approval.
Logic in one sentence, laid out in the Flow Editor.
When a request is ambiguous, it asks back, with answer options.
Example illustration
Mark instead of typing IDs. “These three” is context enough.
It proposes rules with tolerances, you confirm, then it checks.
Hundreds of properties at once. One Ctrl+Z reverts it all.
If a request is ambiguous, it asks instead of guessing.
- An admin switches the AI Assistant on for the project
- Set the allowed models in the Portal
- Open the “AI Assistant” panel in the Editor
From sentence to change
Propose first.
Then approve.
The assistant shows every change as a card first. Only your approval writes it into the project, and one Ctrl+Z takes it back as a whole.
1In the chat
Describe
You say in one sentence what should change.
2In the Editor
Proposal as a card
Every planned change is listed on its own, old and new side by side.
3Your decision
Approve or decline
Once approved, everything lands in the history as one step.
Example illustration
Example illustration
MCP interface · from version 1.32
Your AI.
Your tools.
Every ScaliRo project offers its tools over MCP. Your own AI environment can then work directly in the running project, with the same tools as the assistant in the Editor.
- The same toolboxRead and change the layout, check patterns, create flows, start, pause and stop the simulation.
- A key just for MCPAdmins create API keys of type MCP. They work only on /mcp, expire after one year at most, and act with the rights of the person who created them.
- Off until you switch it on/mcp only answers once the AI Assistant is enabled in the project.
ASCII and binary DXF. Use meters or millimeters as drawing units and set the origin to 0,0. Polylines with varying start and end width are not supported.
- DXF Import › DXF File Requirements
- DXF Import › Importing DXF Files
Example illustration · answer based on the docs page “DXF Import”
Docs AI · with your ScaliRo license
Ask the documentation.
The same AI sits in our documentation as its search. It answers in your words, names its sources, and says when something is not in there.
- Questions in your wordsAsk the way the question comes up in the project, instead of clicking through pages.
- With sourcesEvery answer points to the pages it came from.
- Honest about gapsIf something is not in the documentation, it says so instead of making something up.
Open docs.scaliro.com Sign in with your Portal account
Security
You decide what
the AI may do.
Which models may run is set in the Portal: for the whole organization, a space or a single project.
Off until you switch it on
Assistant and MCP interface stay off until an admin enables the AI Assistant for the project.
Only after approval
Every change appears as a card to approve. One Ctrl+Z reverts it as a whole.
Your rights, nothing more
The assistant only changes what your role allows. Without write access it stays read-only.
No training on your data
Requests go through our own server in Nuremberg and only from there to OpenAI. OpenAI does not use API data for training.
How to get started.
Where you work with the AI
The Portal runs in your browser, with nothing to install. Projects you open there have the AI assistant in the Editor; you set which models are allowed in the Portal.
The Hub is our desktop app: you work on your projects locally on Windows. The AI assistant is coming there next; it will need an internet connection.
Over MCP your client connects to any instance from version 1.32: in the Portal and on the desktop with the Hub, with no extra settings, once the AI assistant is enabled in the project.
Try the AI assistant in your own projects and tell us what is missing. Your feedback shapes how we make building simulations even faster.
In early access: write to us and we will enable it for your project.
Enable the AI Assistant in the project, connect your client. In the Portal with an MCP key, on the desktop without one.
FAQ
Does the AI work out how many robots I need?
No. That stays the job of the simulation. The assistant can start and pause a simulation, but it does not evaluate it. It takes the layout work before that off your hands: checking, aligning, renumbering, building logic.
Does the assistant change anything without asking?
No. Every change first appears as a card with “Approve”, “Allow for this chat” and “Decline”. “Allow for this chat” only lasts until you clear the conversation, close the project, or switch it off above the input.
What is the difference between the assistant and MCP?
The assistant lives in the Editor and sees your selection. With MCP you use the same tools from your own AI environment, such as Claude Code or Codex. That suits developers, scripts and recurring routines.
Can the assistant build logic too?
Yes. It creates flows, adds Flow Units, connects and lays them out. You review the result in the Flow Editor.
Which AI models are behind it?
In the Editor, current models from OpenAI. Which of them may run is set in the Portal per organization, space or project. With MCP you bring your own AI.
Is my project data used for training?
No. Requests go through our own server in Nuremberg and only from there to OpenAI. OpenAI does not use API data for training.
What if an answer is wrong?
It can happen, which is why it says so below the input: check anything that matters. With thumbs up or down you tell us what was missing.
Your layout.
Your questions.
Already working with ScaliRo? Write to us and we will enable the assistant for your project in early access.