Paper Crane is now in beta.
We are opening Paper Crane to a small group of beta users today. It is a catalog of protein design tools — the same models you would run yourself if you had the time, the GPUs, and the patience to wire them together. We did the wiring. The patience is still on you.
The thesis
In Japan, senbazuru is the practice of folding one thousand origami cranes to grant a single wish. The folds are nearly identical. The wish is singular. The discipline is in the repetition.
Protein binder design is the same shape of problem. You do not get one binder from one seed. You get a binder from a thousand seeds, filtered down to ten, validated down to one. The math is brutal and the math is the point. We named the company after the practice because we wanted the name to be honest about what the work actually is.
Why distribution matters
The models that matter — RFdiffusion, BindCraft, BoltzGen, Boltz-2, Chai, ProteinMPNN, ESM-2 — are open. The papers are public. The weights are downloadable. And yet most working scientists still cannot run them, because running them means owning a GPU cluster, speaking three flavors of CUDA, and reconciling a dozen incompatible Python environments.
One company, Tamarind Bio, has done a respectable job stitching the models behind a UI. We respect the work. We disagree with the shape: a walled garden where the catalog is curated and the prices are quiet. Our bet is the opposite — an open catalog, posted prices, your own keys, your own compute, your own data. We make money on the orchestration. You keep the science.
A thousand folds for one wish. A thousand seeds for one binder. The patience is the product.
What is in the beta
Twelve tools, wired end to end: backbone generation, sequence design, structure validation, ranking, and quality control. Bring your own Modal account and we run jobs on your GPUs at cost — no markup on compute. A public benchmark, updated nightly, so you can see how each tool performs on standard targets before you spend a single GPU-hour. A typed CLI for the people who live in terminals. A REST API for the people who do not.
There is no free tier and there is no enterprise tier. There is one flat seat price during beta. Everyone gets the same catalog.
What is coming
After beta we will add a pharma tier — audit logs, SSO, on-prem deployments, and the paperwork that regulated environments need. General availability is targeted for late summer. Between now and then we will keep adding tools as the field publishes them, and we will keep the benchmark honest.
How to join
Beta is invite-only and small on purpose. If you have a target you are working on and you want a partner who will not lock you into their stack, request access on the pricing page. We read every request.
Fold carefully. We will see you in the catalog.