Node Graphs

A node graph lets you build a part by wiring boxes together instead of stacking operations in the scene tree. To the rest of your design it is just an object: it has one row in the scene tree, you move, colour, export and undo it like any other part, and its settings show in the Properties panel. The only difference is the Show Graph switch, which opens its graph under the 3D view.

Use a node graph when you want to turn a few parts into one tool with a handful of settings - a bracket with a Width and a Hole Size, say - that you can reuse and share.

Start a Graph

Select one or more parts.
Click Convert to Nodes on the toolbar (next to Group and Ungroup).

The selection becomes one object named Geometry Nodes, and its graph opens. Each part you selected is a node, wired to the Output, so the object looks exactly as the parts did.

Show and Hide the Graph

Select the object and turn Show Graph on or off in the Properties panel. With it off you still use the object’s settings, you just don’t see the wiring. The choice is saved with the design.

The Node Editor

Input (on the left) holds everything that comes into the graph: its settings (parameters), the objects it works on, and the Scene and Parameters sockets used by formulas.
Output (on the right) is what the object shows. Whatever you wire into it is the result.
Cards are the nodes in between. Each card is an object - a shape, an operation, a sheet - and shows the same settings the Properties panel would. Click a card to select it; drag its title bar to move it.
Sockets are the dots on a card’s edges: inputs on the left, outputs on the right. Their colour says what they carry - green for solids, blue for paths, grey for numbers, lilac for yes/no values, indigo for positions and sizes (an X, Y and Z together), orange for 3D curves, plum for images, pink for Variables. A single number or yes/no is drawn as a short bar instead of a dot. Hover a socket to see its name.
Diamond sockets are selections: they carry an answer for each edge or corner, not one value. See Choose Edges with Selection Nodes below.
Noodles are the wires. Drag from an output socket and drop on an input socket of the same colour to connect them. To move a noodle, pick it up by the end at its input and drop it somewhere else; drop it on empty space to remove it. A noodle that would make a loop is refused, and so is one between sockets that don’t match - the editor tells you why, for example Types don’t match: geometry into image. A noodle into a diamond socket is dashed.

Scroll the mouse wheel to zoom, and drag with the right or middle mouse button to pan.

Add and Delete Nodes

Press Shift+A, or right-click on empty space, to open the Add Node menu at the pointer. It lists Shapes, Operations, Math (Separate XYZ and Combine XYZ), Sheet, Selection (Position, Compare and Inside Box) and My Tools; type to search. The node you pick lands at the pointer, unwired and selected.
With a graph shown, the toolbar’s operation buttons add a node to the graph you are looking at.
Drag an object from the scene tree onto the editor to move it into the graph, or drag an item from the Library to add a copy.
To delete a node, select it and press Delete or Backspace, or right-click its card and choose Delete Node. Its noodles go with it. Input and Output can’t be deleted.

Every one of these is a single undo step.

Operations as Nodes

Most toolbar operations are also nodes. Wire a shape or an outline into an operation’s Source and wire its Result on to the next node or the Output. A few operations have their own inputs:

Translate, Rotate, Scale and Fit to Bounds take whatever you wire in, solid or outline, and give back the same kind of thing, so Circle -> Translate -> Linear Extrude works. Rotate turns the parts about their own centre, Scale grows them from their bottom centre, and Fit to Bounds fits them in a box centred on them - so the node keeps working when what is wired in changes size or moves. Wire a parameter into a Scale’s Width, Depth or Height to size the part to that number.
Subtract has a Keep input (the parts that stay) and a Remove input (the parts cut out of them). Subtract & Replace has Keep and Replace: the Replace parts are cut from the Keep parts and left in the hole. Switching a node between the two keeps its noodles.
Loft has one Sections input that takes as many outlines as you like. They are stacked by height, lowest first, so the order you wire them in doesn’t matter. To raise a section, wire it through a Translate first.
Sweep has a Profile (the outline to carry) and a Rail (the path to carry it along). Only a 3D Curve’s result - the orange socket - fits the Rail. With nothing wired into the Rail, the sweep follows the S-shaped rail it was made with.
Image to Path takes a picture through its plum Source socket. Drag an image from the Library or the scene tree into the graph and wire its result there; Image to Path traces the outline, which you can then wire into Linear Extrude like any other outline. Only an image fits this socket. (The Image Converter is not a node; it stays a shape you drop in from the Library.)
Bounds takes parts in and gives back a box that fits them, plus the box’s Min, Max, Size and Center as values (see Values Between Nodes below).
Fillet and Chamfer as nodes round (or cut) every sharp edge of what is wired in, since a card has no 3D view to click edges in. Fed an outline, they round every corner. To round only some, wire a selection into their Selection diamond (see Choose Edges with Selection Nodes below). Unchecking Selection, with nothing wired in, rounds none. The toolbar’s Fillet and Chamfer still round only the edges you click.

Values Between Nodes

Some nodes give values as well as, or instead of, a part: a Bounds gives its Min (low corner), Max (high corner), Size and Center, and a Measure Tool gives its Distance. Wire one of these outputs into a setting on another card and the setting follows it every time the graph rebuilds. The card shows the value the last rebuild worked out, in your units.

Min, Max, Size and Center are each an X, Y and Z together (the indigo socket). To use just one number:

Separate XYZ (Add Node > Math) splits a position or size into its X, Y and Z. For example, wire the Bounds’ Size into Separate XYZ’s Vector and its Z into a Cube’s Height to make the cube exactly as tall as the parts in the Bounds.
Combine XYZ does the reverse: three numbers in, one position or size out.
In a formula, add the letter: =Bounds.Max.Z is the top of the Bounds box, and =Bounds.Size.X its width. Inside a graph, wire the Bounds card’s Variables socket to the card whose formula reads it, as for any formula (see Formulas Inside a Graph below).

Choose Edges with Selection Nodes

A Fillet or Chamfer node rounds every sharp edge until you tell it which ones. You tell it with a selection: a question the fillet asks about each edge (or, for an outline, each corner), such as “is this edge above 5 mm?”. Edges that answer yes are rounded; the rest are left sharp.

Selections travel through diamond sockets on dashed noodles. A diamond’s output only goes into a diamond input - otherwise the editor says A field only goes into a diamond socket. A diamond input with a dot in it also takes a plain value, which then answers the same for every edge.

The selection nodes are in the Add Node menu’s Selection group:

Position gives each edge’s own position (for a straight edge, its middle). Wire it through Separate XYZ to test just its X, Y or Z.
Compare tests A against B: Greater Than, Less Than, Equal To, At Least or At Most. Choose Within Distance instead to select the edges within a Distance of a Point. Numbers are in your units.
Inside Box selects the edges inside the box from Min to Max. An edge lying on the box’s surface counts as inside.

Example: Round Only the Top Edges

Add a Cube and a Fillet node, and wire the Cube into the Fillet’s Source and the Fillet into the Output. Every edge is rounded.
Add Position, Separate XYZ and Compare.
Wire Position into Separate XYZ’s Vector, and Separate XYZ’s Z into Compare’s A.
Set Compare to Greater Than and B to the height of the Cube’s middle (10 for a 20 mm Cube sitting on the floor).
Wire Compare’s result into the Fillet’s Selection.

Only the edges higher than the middle - the four top edges - are rounded. The upright edges stay sharp, because their middles are not higher than B.

Example: Round the Edges Inside Another Part’s Box

Wire the part whose edges you want to round into a Fillet node.
Wire a second part - one that overlaps the corner you want rounded - into a Bounds node.
Add Inside Box. Wire the Bounds’ Min into its Min and the Bounds’ Max into its Max.
Wire Inside Box into the Fillet’s Selection.

Only the edges inside the second part’s box are rounded. Move or resize the second part and the rounded edges follow.

Example: Round the Corners on the Right of an Outline

Drag a Box from the Library’s 2D Shapes into the graph. Wire it into a Fillet node, and the Fillet into a Linear Extrude. Every corner is rounded.
Add Position, Separate XYZ and Compare. Wire Position into Separate XYZ, and its X into Compare’s A.
Set Compare to Greater Than and B to an X between the left and right sides of the outline (0 for an outline centred on the origin).
Wire Compare into the Fillet’s Selection.

Only the corners on the right are rounded. To round the corners near one spot instead, set Compare to Within Distance and type the spot into Point.

Parameters

Parameters are the settings the object shows in the Properties panel.

Show the graph.
Drag from the Input node’s empty socket (labelled Drag to a setting or a shape input) to a setting on any card - a Box’s Width, for example.

A new parameter appears on the Input node and as a row in the object’s Properties panel. Change it there and everything wired to it follows. Until you add one, the Properties panel says No parameters yet. Show Graph, then drag from the Input node’s empty socket to a setting.

Rename a parameter by double-clicking its name.
Reorder parameters by dragging the grip at the left end of a row.
Remove a parameter by removing its last noodle; the row goes with it.
A parameter can hold a formula, like =A1 from a Variable Sheet or =Handle.Width from another object.
Other objects can read it too: =Bracket.Width reads the Width parameter of the graph named Bracket. Nodes inside a graph can’t be read from outside; if you try, you see Nodes inside a graph can’t be used from outside it; add a graph parameter instead.

Object Inputs

A graph can also take objects in, so it works like an operation. Drag from the Input node’s empty socket to a card’s shape input (the socket an operation reads its source from). This adds an object slot to the Input node. Objects you converted with Convert to Nodes are already wired in this way.

Formulas Inside a Graph

A graph is sealed: a formula inside it sees only what is wired to its card’s Variables socket. If a formula names something that isn’t wired, the setting shows a warning telling you which noodle to add:

To read another node, like =Box.Width, drag from Box’s Variables socket to this node’s Variables socket. Otherwise: Box isn’t wired to this node. Drag from Box’s Variables socket to this node’s Variables socket.
To read the graph’s own parameters, like =Input.Width, drag from the Input node’s Parameters socket to this node’s Variables socket. Otherwise: Input isn’t wired to this node. Drag from the Input node’s Parameters socket to this node’s Variables socket.
To read objects in the scene outside the graph, wire the Input node’s Scene socket to this node’s Variables socket. Otherwise: … isn’t wired to this node. Wire the Input node’s Scene socket to this node’s Variables socket to use names from the scene.
A Sheet node works the same way: wire its Variables socket to each node whose formulas read its cells. A sheet has no result, so it never goes into the Output.

Save as Tool and My Tools

Right-click the object and choose Save as Tool to keep a copy you can use in any design. Tools are stored in My Tools in the Library.

A tool that makes a shape (no object inputs) is also added to the Favorites bar. Drag it into your design like a primitive.
A tool that works on objects (it has object inputs) gets a button in the toolbar’s My Tools group. Select a part and click the button to run the tool on it.

Each use adds an independent copy, so changing one never changes the others. Tools also appear under My Tools in the Add Node menu.

Graphs Inside Graphs

Add a graph to another graph - from the Add Node menu’s My Tools, the Library, or the scene tree - and it becomes one card. Its sockets are the inner graph’s parameters and object inputs, plus a Result. Leave an object input unwired and the inner graph keeps its own object.

To edit the inner graph, double-click its card or click Open. The breadcrumb above the editor (for example Outer > Inner) shows where you are; click it or press Escape to step back out. Changes inside affect only that copy.

Graphs can be nested up to 8 deep. Deeper than that, the graph shows Graphs can be nested at most 8 deep. Apply or remove a graph inside this one to build it.

Fillet - Rounding edges, and picking them in the 3D view
Bounds - A box that fits other parts, and its Min, Max, Size and Center
Components - Another way to show just a few settings for a group of parts
Expressions - Formulas in settings
Object References - Reading another object’s settings with Name.Property
Variable Sheet - Shared values for a design