Start here
The single biggest lever on output quality isn't which model runs — it's what you put in the brief. Five things, in order of how much they move the result.
The Design agent turns your brief into candidate geometry. The DFM agent then checks that geometry against a real rulebook — wall thickness, overhang angle, hole size, and more, per process. If your brief doesn't say what process, what scale, or what matters most, the agents have to guess. A guess produces a generic critique. A specific brief produces a specific one, with real numbers attached.
FDM, SLA, 3-axis CNC, injection moulding or sheet metal — each has a different rulebook (a CNC part can't have an undercut; an FDM part can, but pays for it in overhang support). If you don't know yet, say what you're optimizing for (cheapest per-unit at low volume, tightest tolerance, best surface finish) and let the brief mention that instead of guessing a process.
“200 units” and “under $8/unit” changes the cost model's amortized setup cost, which changes whether a design that's cheaper to make but more expensive to tool actually wins. A brief with no volume gets a wide, low-confidence cost range — because at 10 units vs. 10,000 units, the right answer can flip entirely.
Every part has one thing that, if violated, kills the design — a gasket seal, a snap-fit tolerance, a weight limit, an aesthetic rule like “no visible fasteners.” Say it explicitly. The DFM agent weighs every finding against the rulebook regardless, but a stated constraint becomes something the Design agent actively designs around instead of something DFM has to catch after the fact.
Even a loose one (“PETG or PLA+”, “aluminium”) narrows the Material Advisor's shortlist to trade-offs that matter for your part, instead of a generic ranking.
A text description of wall thickness is a claim. A measured mesh is a fact. See uploading an STL — even a rough block-out is worth uploading before you invest in a polished model.
FDM · 200 units · $8/unit
“Design a modular desk caddy for a small studio: phone stand, two pen cups, and a tray for USB-C dongles. Must print on a consumer FDM printer without supports if possible. Target unit cost under $8 at 200 units. Aesthetic: calm, slightly industrial, no sharp user-facing edges.”
Process, volume, budget, the constraint that matters (“no supports”), and an aesthetic rule — all in four sentences.
3-axis CNC · 300 units · $30/unit
“Design a weather-resistant enclosure for an outdoor air-quality sensor board (80 × 55 × 30 mm). Needs an IP65-style gasket seal, four M3 mounting bosses, a cable gland on one end, and a removable lid. Machined from aluminium for a 300-unit pilot. Target unit cost under $30. Aesthetic: clean, anodised, no visible fasteners on the front face.”
Exact board dimensions, the seal requirement that drives the whole design, fastener count, and a material — a CNC-specific brief a generic one would never produce.
“Make me a phone stand.”
This still runs — the agents fill every gap with a reasonable default and say so in the run's confidence basis. But the DFM critique will be generic (principle-level advice, not a measured violation), the cost range will be wide, and the manufacturer match will be a guess. None of that is a bug; it's the honest result of an under-specified input.
Want to try this against your own part right now, no account needed? Run the free instant check.