2026.09.01 2026.09.01

Generate Gridfinity Bins Instead of Hunting for Them — Four Kinds of Parametric Tools

swiftwand

Search a model site for Gridfinity bins and thousands of results come back. Thirty minutes later you have twenty tabs open and have not started a single print. Searching costs more than people expect. And even when the width is right, the depth routinely is not.

There is another way through this. Enter the dimensions and generate the bin on the spot. Gridfinity bins have a shape fixed by the standard, so given width, depth, height and a divider count, the geometry follows mechanically. Parametric generation replaces the job of finding a file with the job of deciding numbers.

There are roughly four families of generator. This article sorts out what each is good at, and the conditions people overlook when choosing.

When generating beats searching
The four families
The OpenSCAD route
The CAD add-in route
Staying in the browser
Why asking a model to write the code fails
Three checks before printing
Summary
Sources

忍者AdMax

When generating beats searching

Generation is not always the better move. The dividing line is simple.

Situation	Better approach
You want a plain divided box	Generate. Searching takes longer than making it
You want a cradle shaped to a specific tool	Look for an existing model, then design one if none exists
You want to absorb the remainder of a drawer	Generate. Half-unit sizes are available
You want a decorative box that looks good	Existing models

If the content of the box is space, generate. If the content is a shape, search. A divided box you throw screws and drill bits into is the former; a cradle following the outline of a ratchet handle is the latter.

The time comparison is worth making explicit. Searching, comparing, checking dimensions and downloading routinely takes twenty to thirty minutes, and settling for something that roughly fits leaves a visible gap after printing. Generation, once the tool is installed, is entry of numbers and a few minutes to export. The setup cost lands once and drops to near zero afterwards, which suits storage precisely because storage repeats.

Speed is not the only benefit. Save the settings and you can add an identical box six months later. Published files sometimes disappear; your settings do not.

The four families

Each family assumes something different about your setup. Setup weight and flexibility run roughly inverse to each other.

Family	Where it runs	Setup weight	License to check
OpenSCAD scripts	Desktop, OpenSCAD	Medium	MIT or GPL-3.0 depending on the project
CAD add-in	Desktop, commercial CAD	Heavy	CC BY-NC-SA 4.0 plus the CAD license itself
Browser generators	Nothing installed	Light	Varies, check per tool
Photo to outline	Container on your own machine	Medium	MIT

The OpenSCAD route

The most established option is a script-based generator you drive through a customizer. The best-known one is MIT licensed, which makes it the least encumbered starting point of the four.

Item	What it offers
Base dimensions	42 mm base width, 7 mm height unit
Baseplate styles	Thin, weighted, skeletonized, and screw-together variants
Holes	Togglable magnet holes for 6 mm diameter by 2 mm magnets and M3 screw holes, with countersink or counterbore options and a corners-only mode
Bin options	Stacking lip, scoop, label tabs, and solid generation

Solid generation is the quiet strength of this family. To build a cradle shaped to a tool, you can emit a solid block first and subtract the tool shape from it. That gets you there faster than designing a box from nothing, and it never leaves the standard.

A second script-based project exists under GPL-3.0 with a broader option set. The license difference matters if you plan to redistribute your own modified version, since a copyleft license carries obligations that MIT does not.

The CAD add-in route

If you already run commercial CAD, an add-in generates bins inside the tool you know. The advantage over the script route is what happens after generation: the CAD history still applies, so you can add a fillet, drill a hole, or thin a section afterwards. The generated body is a starting point rather than a finished export, which is the decisive difference against browser generation.

There is a condition. The add-in is licensed CC BY-NC-SA 4.0, which includes a non-commercial clause. A price of zero on the listing page and permission to sell the output are separate questions. The CAD application itself needs its own license on top.

Staying in the browser

If you want to install nothing, browser-based generators exist. One caution first: the site that collects Gridfinity tools is a directory, not a generator. What it gives you is a curated list.

The tools you reach from it differ in character. One extracts the outline of an object from a photograph and cuts a matching recess into a bin; it is published under AGPL-3.0. Another is built on CADQuery and can export STEP as well as STL.

That export format matters more than it looks. STL approximates surfaces with triangles, which is fine for printing but poor for later edits like widening by 3 mm. STEP keeps the surface and curve definitions, so CAD can reopen it and change dimensions. One-and-done, take STL; expecting to revise, take STEP. Checking the export format while choosing the tool saves a rebuild later. The tolerance side of revising a fit is covered in the article on tolerance and fit.

Browser generation has another advantage: the same result from any machine, with nothing to reinstall. The trade is that you depend on someone else keeping the site up, so export and keep the files that matter.

Why asking a model to write the code fails

Asking a language model to write code that produces a Gridfinity bin from scratch tends to fail, and the reason is simple: there is no guarantee the standard dimensions inside the model are correct.

Gridfinity has no specification document maintained by its creator. The numbers are carried by distributed files and community implementations, and documented by the community rather than the author. That is a domain where training-data accuracy is not underwritten, and a plausible-looking number is not verifiable at a glance. Getting the grid pitch right while the rim profile or hole positions drift is enough to destroy compatibility.

So invert the workflow. Supply the standard dimensions yourself as a specification, and let the model do assembly and modification. If a verified implementation already exists, building a diff on top of it is faster and safer than starting over.

Think of it as verification cost. Code written from scratch requires you to check the grid, the rim, the holes and the stacking behaviour yourself before you can trust it. A diff against an existing implementation only requires checking what you touched. Same output, an order of magnitude less to verify. Fast generation with slow verification is not fast. The same framing appears in the article on driving OpenSCAD with a language model.

Three checks before printing

Unit count of the exterior. Divide the drawer interior by the grid and decide where the remainder lands
Hole handling. If magnet or screw holes are enabled, confirm the shape prints without supports
The stacking lip. Disabling it means the bins no longer stack, which saves material for standalone boxes

Only the first of these is beyond the generator, because measuring the drawer is a human job. Which also means that once you have the numbers, the rest is entering them. The measuring itself is covered in the article on measuring drawers and equipment.

Decide the hole question early. Magnets stop bins from shifting when the drawer opens, at the cost of the magnets themselves and the labour of embedding one in every box. Across dozens of bins that labour is not negligible, so putting magnets only in the shallow drawers that actually move is the realistic compromise. The same logic applies to the lip: adding one to a box you never intend to stack costs material and height for nothing.

Summary

Generating Gridfinity bins beats hunting for them more often than not. To recap:

The rule of thumb is space versus shape. A box holding space should be generated; a box holding a shape should be searched for
Four families: OpenSCAD scripts, a CAD add-in, browser generators, and photo-to-outline. Setup weight and flexibility run inverse
Licenses differ per tool. MIT, GPL-3.0, and CC BY-NC-SA 4.0 all appear among free generators, and the non-commercial one blocks selling the output
A CAD add-in lets you keep editing the generated body afterwards, which browser export does not
STEP export matters if you intend to revise dimensions later; STL does not survive that well
Do not have a language model write the standard from scratch. Supply the dimensions and let it produce a diff against a verified implementation

Once you can produce bins on demand, the next constraint is printing dozens of them, which is the batch printing article. Choosing the standard in the first place is the grid standards comparison.