Introduction
User Interface
Components
Scripting
Framework
Physics
Virtual Commissioning
Digital Shadow
Analysis
Tutorials

Torq

Torq is our AI assistant and agent that can help you create models in ProtoTwin. Some examples of things that Torq can help with are:

  • Explain how to create entities and configure components.
  • Explain how to use the TypeScript and Python APIs.
  • Create scripted components, tools and value converters.
  • Create scripted robot programs that are executed by the integrated robot controller.

Help Assistant

We support a free AI assistant on our documentation page. The AI assistant can answer many questions about ProtoTwin, such as:

  • How do I model a SCARA robot?
  • How do I simulate a conveyor?
  • How do I disable collisions between two rigid bodies?
  • How do I change the color of an entity?
  • How do I connect ProtoTwin to a Beckhoff PLC?
  • How do I define a linear relationship between two joints?
  • How do I turn a light on when a button is pushed?

The AI assistant has access to every page on our documentation, allowing our assistant to retrieve relevant information and respond to your query quickly without requiring direct communication with our support team. Please note that the AI assistant on our main website cannot execute code or provide code in its answer to your query. If you need Torq to generate code, you must ask the AI agent in the main application.

AI Agent

Torq, integrated into the main application, supports an agentic workflow for digital twin development. This allows Torq to:

  • Read, modify and create scripted components, tools, value converters and robot programs.
  • Read and modify the contents of your model.
  • Create entities and add components to entities.
  • Connect component IO properties to other component IO properties.

Tools

Torq is provided with a set of tools that enable the agent to solve complex problems:

  • The HELP ASSISTANT tool allows Torq to query a subagent that searches the ProtoTwin documentation for relevant information. Calling this tool is equivalent to asking the assistant on our documentation page.
  • The RUN SCRIPT tool allows Torq to execute a standalone script, which is executed by ProtoTwin at design time. The standalone script can be used to gather information about your model or make modifications to your model.
  • The other tools allow Torq to retrieve examples of scripted components/tools, TypeScript API definitions, list/read/write/delete/rename files and detect errors. These additional tools are similar to the tools the agent integrated into VSCode and Cursor are provided with.

Capabilities

These agentic capabilities allows Torq to fix issues with your scripted components and robot programs, understand the context and purpose of the model you are building, and increase the rate of development of digital twins. Torq has access to the ProtoTwin TypeScript API and the context of your model, meaning Torq may execute tasks such as:

  • Replace the aluminum graphics materials in my model with the stainless steel material.
  • Explain why when the distance sensor detects a box, the conveyor is not powered off.

Torq can accomplish the first task by executing the RUN SCRIPT tool and write code to replace all instances of one material with another material. Since Torq can also read the contents of your model, it can determine whether the logic inside a scripted component is correct or not, determine if you added the component to an entity in the scene, and determine whether the handles correctly reference the sensor and conveyor motor to answer the second question.

Attachments

You can attach PDF files and images to your query to provide Torq with additional context. It may be useful to attach an image if you want Torq to create a visual 3D mesh from the image, or so that Torq can see an image of the model on your screen and resolve issues that are difficult to interpret with natural language.

Recommendations

You can ask Torq follow-up questions if the generated code does not work the first time, or if the query was misunderstood. In general, Torq performs best when you provide detailed instructions or a detailed specification of requirements with a clearly defined scope. Avoid making open-ended or vague requests, such as: “write a component to control my gantry” or “control my machine for me”.

Permissions

You can change the permissions of Torq in the model settings. You can prevent Torq from reading or writing files or executing scripts using the RUN SCRIPT tool. Disallowing these permissions allows you to use our AI assistant with a non-agentic workflow. Please consider that if you prevent Torq from reading the contents of a script, it will not be able to answer questions such as “Can you modify the script to reference a color sensor instead of a distance sensor” or “Fix the error in the Controller.ts file”.

Likewise, if you disallow write permissions for scripting, Torq will not be allowed to create a file so questions like “Create a scripted component called Controller that turns a conveyor off when a sensor is blocked” will cause Torq to output the code in its text response without automatically creating the file for you. You will then have to create the file manually through the script editor, and copy and paste the code to use it.

Non-Agentic Workflow

You can ask Torq in the main application to generate short code snippets for you, such as:

  • How do I reparent an entity?
  • How do I find all the top-level entities in the model?
  • How do I turn a conveyor off when a sensor is blocked?
  • How do I create a sequence that commands a motor to cycle through a list of positions?

If you ask Torq questions in this manner, it will be inclined to output code snippets in response to your query rather than create/modify a file for you. Likewise, if you disable write permissions, Torq will be forced to output code in this manner.

Agentic Workflow

You can ask Torq in the main application to execute scripts, create or read files and apply modifications your model, such as:

  • Describe the contents of the model and explain what the model does.
  • Create a tray with rounded corners.
  • Create a component that animates an entity up and down.
  • Create a tool that applies a force to a rigid body with physics.
  • Create a scripted robot program that alternates moving between two target entities.
  • Add the scripted robot program to the top-level entity and call it from the main program on the robot controller.

Examples

Below we demonstrate the capabilities of Torq with examples of increasing complexity:

1. SCARA Robot Pick and Place

I have a model of a SCARA robot in the scene. Can you add a robot controller and configure it by adding instructions to the main program in order to peform a pick-and-place. There is a sensor that is blocked when a box is ready to be picked.

2. Euro Pallet with Boxes and Slip Sheets

Create a detailed euro pallet, with physics. Add 3 layers of boxes, separated by slip sheets. Add sensible graphics components to everything.

3. Sick DT35 Distance Sensor Component

I’ve attached the manual for a Sick DT35 distance sensor. Can you create a component for this distance sensor, with matching IO. Use the physics query raycast to measure the distance. Ensure that the IO matches that in the manual.

Pricing

As mentioned earlier, the AI assistant on our documentation page is completely free to use. For Torq integrated into the main application, the cost is determined based on token usage. Every new ProtoTwin account is given 2 credits as a welcome bonus. You can purchase additional credits to use our AI agent more often. You will also be assigned credits based on the license you have purchased:

ProtoTwin License
Motion Simulate Connect
Credits 10 50 100

The credit cost of a conversation with Torq is determined based on token usage. If you wish to use a more advanced model such as GPT 5.5 or Claude Opus 4.7, the cost of a query will be much greater than using a cheaper model such as DeepSeek 4 Flash or Gemini 3 Flash. Increasing the reasoning effort will also cause credit usage to be higher.

Models

ProtoTwin currently supports the following AI models:

  • DeepSeek 4 Flash
  • Gemini 3 Flash
  • GPT 5.4 Mini
  • Gemini 3.1 Pro
  • GPT 5.5
  • Claude Opus 4.7

Suggestions

Please refer to the pricing from the model providers and benchmarks to determine which model is most suitable for your task. Provided below are some example queries with the models we recommend:

  1. Create a stack of 10 cardboard boxes
  2. Connect the gripper signal on the robot controller to the state of the suction gripper.
  3. Explain how to turn the conveyor off after a sensor has detected a pallet.

DeepSeek 4 Flash, Gemini 3 Flash, GPT 5.4 Mini are usually the most appropriate since they can accomplish the task perfectly with the smallest cost.

  1. Create a detailed model of a yellow forklift with physics joints and motors for the lift and wheels. Allow me to control the forklift using the arrow keys and create a pallet of wooden boxes on the floor with physics. I want to transport the pallet using the forklift.
  2. Create a program for a 6-axis robot arm that picks parts off the first conveyor and stacks them on the second conveyor. After the 2 parts are stacked, the second conveyor should be turned on for a short period of time to make space for the next stack.
  3. Create a component to control a turntable conveyor. The turntable must rotate 90 degrees and alternate between sending pallets to the left and right to lower congestion. The turntable conveyor motor should be powered off 1 second after the input sensor is blocked. The turntable should rotate back to 0 degrees after either the left end sensor or right end sensor is cleared.

Consider using GPT 5.5, Claude Opus 4.7 or Gemini 3.1 Pro when the task is more complex, requires calculation, high reasoning effort or spatial intelligence.