Two modes for two kinds of cuts. Auto Nest runs a Rust pipeline with sparrow-style Guided Local Search to optimize roll yield in seconds. Free-roam gives operators a manual canvas when the spec demands it.
The real engine packs in under three seconds for typical sheet sizes. The animation below cycles every seven seconds — same parts, same logic, same Rust code path.
The packing pipeline is a real Rust workspace — four crates, full Cargo test coverage, and a CLI binary the Next.js app shells out to via stdin/stdout JSON.
The geometry layer is jagua-rs, a research-grade polygon nesting library. It handles non-rectangular parts, rotation, and concave shapes — not just rectangles.
A sparrow-style GLS metaheuristic explores the placement space, escaping local optima the greedy heuristic can't. The result is yields that beat hand-planning by double digits.
A standalone Rust binary that takes a JSON pack request on stdin and writes the layout on stdout. Same binary runs locally, in CI, in the production server, and in a future native desktop build.
Every Auto Nest run records its full layout JSON to cut_history. The exact placement of every part, the yield percent, the operator, and the timestamp — replayable forever.
Some cuts don't optimize cleanly — a defect on the roll, a customer-specific orientation, a rush job. Free-roam gives the operator a live canvas with snap, rotate, and overlap detection while preserving the same audit trail Auto Nest produces.
The engine is benchmarked against typical sheet sizes (4–9 sqm) with 8–24 parts. These numbers are measured on a single c5.large equivalent.
Per-seat pricing. 14-day free trial, no credit card. Run real cuts against your own stock — if it doesn't pay for itself the first week, walk.
See pricing