Powered by
on
AI processors
High Throughput Low Cost Drug Discovery
Structure prediction, binding affinity, de-novo design, and protein embeddings. Every model, free to try.
¥0
Free demo API, no key.
100%
Japan Hosted
1:1
accuracy parity with the GPU implementations, measured.
How to use it.
01
Workbench
Fold and design in the browser. No install, no signup.
Open the workbench → 02 APIOne HTTP endpoint for every model.
POST api.japanfold.com/v1/predictions
Read the API docs →
03
Agent skill
Ask in plain language. Claude Code, Cursor, Codex and more.
npx skills add moritztng/japanfold
Install the skill →
Why
Fast and cheap because every model runs on Tenstorrent AI processors instead of GPUs, with the whole stack built as one, from silicon to serving.
Models
Every model, running today.
StructureBoltz-2Structure + binding affinity
StructureESMFold-2Rapid single-sequence folding
StructureProtenix-v2All-atom complex prediction
StructureOpenDDEAntibody–antigen complexes
DesignBoltzGenDe-novo binder design
DesignRFdiffusion3Backbones, binders, scaffolds
EmbeddingsESMCProtein language embeddings
EmbeddingsSaProtStructure-aware embeddings
Contact
Talk to us.
Questions, pilots, partnerships: we read every message.