Neurocad user guide

Neurocad is an execution layer between your static engineering files and your EDA and CAD applications. You bring in a file — a datasheet, a drawing, a schematic — tell Neurocad what to extract and how to transform it, and the output goes directly into Altium, Cadence, SolidWorks, or whichever tool you're working in. No manual re-entry. No redrawing from scratch.

This guide walks you through setup, your first recipe, and the building blocks you'll use from there.


How Neurocad works

Every Neurocad workflow follows the same loop:

  1. Bring in a file — a PDF datasheet, engineering drawing, image, schematic, or similar static asset
  2. Build a recipe — a visual workflow that tells Neurocad what to extract, how to interpret it, and what format to output
  3. Run and review — Neurocad synthesizes the data and surfaces the result as a digital artifact you can inspect before sending anywhere
  4. Send to your EDA and CAD applications — push the output directly to your connected application

Recipes are reusable. Once you've built one for a footprint extraction or a CAD part from a drawing, you can run it again on new files or automate it for batch processing.


Key concepts

Recipe A recipe is a visual workflow built in the Canvas editor. It defines the inputs (your files), the logic (your prompts and processing nodes), and the outputs (the digital assets going into your EDA and CAD applications). Recipes are the primary unit of work in Neurocad.

Canvas editor The visual workspace where you build and edit recipes. Nodes are connected left to right — data flows from your input file through processing steps to the output.

Node A single step in a recipe workflow. Each node has a specific function: adding a file, extracting geometry, processing library data, sending to a connected tool, and so on. Nodes are selected from the Node palette and connected in the Canvas editor.

Preset recipe A pre-built recipe scaffold for common workflow operations. Presets give you a working node structure to start from so you're not building from scratch. You can customize any preset inside the Canvas editor.

Artifact The output of a recipe run — a footprint, symbol, CAD part, schematic, or other digital asset. Artifacts are reviewable before you send them to any connected tool.


Step 1: Setup

Download and install Neurocad Link

Follow the download instructions provided at installation for Neurocad Link. Once installed, sign in on Neurocad Link. Then you’re done. 

Neurocad Link will auto install into your current suite of tools upon download. Should you have problems, each application has its own connection guide. 

Current supported applications:

  • Altium
  • Cadence
  • Fusion
  • SolidWorks
  • Inventor


Step 2: Open app.Neurocad.io 

Click Start tour to get a quick orientation of the main menu.


Your first recipe

This walkthrough uses the Extract symbol preset — a common starting point that demonstrates the full Neurocad loop.

Open a new recipe

  1. Click Recipes in the left navigation panel, or click the Neurocad icon at the top left for the dropdown menu, then select Recipes.
  2. Click New recipe. You'll land on the Neurocad main menu.
  3. Select Extract symbol from the preset recipe list. This opens the Canvas editor with a pre-built node structure that reads and interprets component library parts.

Add your file

  1. In the Canvas editor, start at the Add file node on the left. Drag and drop your file or select it from your computer. The file will need to be a component library part. The Add file node accepts single page and multipage PDFs and text files, JPEGs, PNGs, and more.
  2. The Extract node will automatically display a visual representation of your file.
    • Multipage document (e.g., PDF or text file): In the Extract node, toggle to the correct page and zoom in on the section within the dotted-line box. 
    • Single page document: Re-center and zoom the reference image so it's fully visible within the dotted-line box.

Write your prompt

  1. In the Prompt node, describe what you want Neurocad to extract from the file. This works like LLM prompt engineering — be specific. If the image has ambiguous labels, measurements, etc., clarify what you're asking the agent to re-create. Instead of "extract the footprint," write something like: "Extract the land pattern and pad dimensions from the component footprint on page 3. Output as a PCB footprint compatible with IPC-7351 naming conventions."

Run the recipe

  1. Click Run on the Library node.

Review the output

  1. Once the recipe runs, review the output in the Library node or open the Artifacts panel (click the Artifacts icon in the Canvas editor) to inspect what was generated. The output will also appear on your Overview dashboard under the Artifacts card.

    Don't skip this step. Reviewing the artifact before sending it downstream is your quality gate. It's faster to correct a prompt and re-run than to clean up a bad import inside your EDA and CAD applications.

Recipe errors

  • If you receive a failure message, or the output is not what you expected, try rewriting your prompt or verify your file type works in the preset recipe. (If you're using a Library preset, the agent will not generate a schematic or model, and vice versa.) For further help, contact us in the Neurocad Slack channel or support@neurocad.com.

Send to your EDA and CAD applications

  1. In the Live design node, use the arrows to toggle to your EDA and CAD applications. Confirm there's a green circle next to its name indicating it's connected, then click Send to desktop. If no green circle appears, check your connection in Neurocad Link, and confirm you're logged into the EDA and CAD applications.

Indicator Status

  • Green - Connected/ready
  • Red - Not running/disconnected
  • No circle - Not yet installed


Step 3: Building custom recipes

Once you're comfortable with presets, you can build recipes from scratch using the Blank recipe option, or modify any preset to fit your workflow.

The Node palette

The Node palette on the left side of the Canvas editor is organized into five sections. Click any node to add it to your canvas, then connect nodes left to right to build your workflow.

Input nodes

Input nodes provide data to your recipe. They appear in the Input section of the node palette.

Add file - Upload any file as an asset. This is typically the starting point for image-based workflows. The Add file node accepts single page and multipage PDFs and text files, JPEGs, PNGs, and more.

Prompt - A text field where you describe what you want to create. Most process nodes accept a prompt input for context.

Describe what you want Neurocad to extract from the file. This works like LLM prompt engineering — be specific.

If the image has ambiguous labels, measurements, etc., clarify what you're asking the agent to re-create. Instead of "extract the footprint," write something like: "Extract the land pattern and pad dimensions from the component footprint on page 3. Output as a PCB footprint compatible with IPC-7351 naming conventions."

Extract image - Takes a file (PDF, SVG, 3D model) and extracts a raster image from it. Useful for pulling reference drawings out of documents before sending them to a process node (for a detailed conversion). Currently available for Model interpretations.

Import - Import EDA and CAD libraries and schematics directly. Outputs a library, schematic, and preview image.

Corner set - Define parameter corners for sweep runs. Outputs both the active corner and the full set of corners.

Connect read - Defines an input port for automation recipes. Used when a recipe is triggered programmatically rather than run manually.


Mechanical nodes

Mechanical nodes run remotely and handle CAD tasks. They appear in the Mechanical section of the palette.

Model - Creates 3D CAD models from drawings and prompts. Accepts separate image inputs for part and enclosure references.

Interpret drawing - Analyzes an engineering drawing and extracts structured dimensions and parameters. Outputs a parameter set that can feed into a Model node for more precise geometry.

Other Mechanical nodes (Drawing, Assembly, Sketch) are planned but not yet enabled.


EDA nodes

EDA nodes run remotely and handle electronic design tasks. They appear in the EDA section of the palette.

Library - Creates component library symbols and footprints from reference images and prompts. Accepts separate image inputs for symbol and footprint references.

Other EDA nodes (Schematic, PCB, CAM, Wire diagram, SPICE) are planned but not yet enabled.


Output nodes

Output nodes are the terminal step of a recipe. They appear in the Output section of the palette.

Live design - Sends the result to your local Neurocad installation for viewing and further editing.

Run recipe - Executes the entire recipe as a job. Also used to publish the recipe as an embeddable link.

Embeddable - Publishes the output artifact so it can be embedded or shared externally.

Connect write - Defines an output port for automation recipes. Paired with Connect read to form the boundary of a headless pipeline.


Preset recipes

When you create a new recipe you can start from a preset template. These wire up common workflows so you can get started quickly.

EDA presets

Extract footprint - File > Extract image > Library (footprint mode) > Live design

Extract symbol - File > Extract image > Library (symbol mode) > Live design

Mechanical presets

CAD part from drawing - File > Extract image > Model (part mode) > Live design

CAD enclosure from drawing - File > Extract image > Model (enclosure mode) > Live design

CAD part from interpreted drawing - File > Extract image > Interpret drawing > Model > Live design. The interpret step extracts dimensions from the drawing and passes them as parameters to the model node.

Automation presets

Automation: Library symbol - Connect read > Library > Connect write

Automation: CAD part - Connect read > Model > Connect write

Automation presets use Connect read/write instead of file inputs and live design outputs. They are designed for triggered or headless execution. The typical workflow is to prototype with a standard preset first, then adapt for automation once the flow works.

Blank recipe

Blank recipe - An empty canvas. Build your recipe from scratch.


How nodes connect

Nodes communicate through typed ports. The main port types are:

Text - Prompt and description strings

Image - Raster images extracted from documents or uploaded directly

Data - Structured data like parameter sets and sweep results

EDA - EDA artifacts (libraries, schematics, netlists)

Mechanical - CAD artifacts (3D models, drawings, assemblies)

Generic - Accepts any type. Used by output nodes so they can receive artifacts from either domain.

The canvas enforces type compatibility when you draw connections. Generic inputs accept any output type, but typed inputs only connect to matching outputs.


Step 4: Automate

Once a recipe is working reliably, you can apply it to an automation sequence for batch processing.

Go to Automate in the left navigation panel. From here you can configure automations that run a recipe across multiple files or trigger on a schedule. Automate is useful for library maintenance, bulk part imports, or recurring drawing conversions.


Navigating the interface

Main navigation

The left panel is your primary navigation:

  • Recipes — Build, edit, and run recipes
  • Connect — Integrate EDA and CAD applications with Neurocad
  • Automate — Run recipe automations
  • Account info — Settings and sign out (bottom left of the dashboard)

Overview dashboard

Click the Neurocad logo or Neurocad icon at the top left to access Overview — your home base for recent activity.

  • Recently visited — Recipes you've recently created or opened
  • Recently run — A log of recent recipe runs with status (completed, failed, in progress)
  • Artifacts — All outputs from your recipe runs, filterable by domain and searchable by name

Canvas editor controls

Use the icon panel at the bottom of the Canvas editor for workspace navigation:

  • Hand icon — Move the workspace
  • Frame icon — Re-center the workspace
  • Microscope icon — Zoom in and out

Recipe menu

Click the caret (v) next to a recipe title for recipe-level options:

  • Rename the recipe
  • Duplicate the recipe
  • Sync (syncs this recipe with our servers for automation)
  • New recipe (returns to Overview)
  • All recipes (returns to the Recipes dashboard)


Reference

Failure messages

If a recipe run fails, start here:

  1. Rewrite the prompt — Most failures are prompt-related. Describe what you want Neurocad to extract from the file. This works like LLM prompt engineering — be specific. If the image has ambiguous labels, measurements, etc., clarify what you're asking the agent to re-create. Instead of "extract the footprint," write something like: "Extract the land pattern and pad dimensions from the component footprint on page 3. Output as a PCB footprint compatible with IPC-7351 naming conventions."
  2. Verify the file — Confirm the file is supported and that the relevant content is visible in the Extract node.
  3. Check your connection — If the failure occurs at the Live design node, confirm your EDA and CAD applications are connected and you're logged in.
  4. Contact support — Connect with us on our Neurocad Slack channel, or via email: support@neurocad.com


Last updated April 10, 2026