01CONTRACTS
TYPED PAYLOADS, NOT PDFS
Every model returns JSON with typed fields and URLs to PDB, mmCIF, or CSV artifacts. No parsing PDFs or screenshots.
MISSIONCOMPUTATIONAL BIOLOGY
BIOLOGY RUNS ON CURIE
Inference infrastructure for biology AI. Run protein folding, molecular docking, genomics, and single-cell models through one API. No GPU management. Pay per call.
ALPHAFOLD 2ESMFOLDPROTEINMPNNDIFFDOCK 2.0NUCLEOTIDE TRANSFORMERSCGPTGENEFORMERBOLTZ-2ENFORMERMOLMIM18 MODELS1 API<200MS LATENCY
MODEL CATALOG
[18 models, 14 live]
/
/
/
/
/
TYPE
NAME
DESCRIPTION
SOURCE
LATENCY
STATUS
ONE API. EVERY MODEL.
import curie
client = curie.Client(api_key="sk-...")
# Protein structure prediction
result = client.run(
model="esm/esmfold-v1",
sequence="MKTIIALSYIFCLVFA..."
)
# Molecular docking
dock = client.run(
model="mit/diffdock-2",
protein_pdb=result.pdb,
ligand_smiles="CC1=CC=CC=C1"
)
print(dock.binding_affinity) # -8.4 kcal/mol< 200ms
median latency
18+
models available
1
unified API schema
OUR MISSION
PRODUCTION INFERENCE FOR BIOLOGY AI
Curie provides API access to 18 biology models across protein folding, molecular docking, genomics, RNA, and single-cell analysis. Run predictions at scale without managing infrastructure.
One API schema. Typed responses. Per-call billing. Models are hosted, versioned, and monitored. You send a request, get back structured data and downloadable artifacts.
LATENCY
Median response under 200ms for most models. Structure prediction completes in minutes.
NO INFRASTRUCTURE
No GPU cluster. No model weights. No deployment pipeline. Call the API and get results.
PRICING
Pay per API call. No subscription. No minimum spend. Scale from prototype to production.
MODEL STACK
WHAT RUNS UNDER THE HOOD
All models return the same JSON schema. Swap model in one field and your code keeps working. Weights are versioned and responses include revision metadata.
SLUGS BELOW ARE THE LITERAL API.MODEL IDENTIFIERS YOU PASS TO /V1/INFERENCE.
COMING SOON
MORE MODELS IN DEVELOPMENT
Fold ensemble, Sequence foundry, Pocket docking, ADT & liability prediction, and Assay bridge coming Q3 2026
PRODUCT PRINCIPLES
HOW IT WORKS
Send a request to one endpoint. Get structured JSON and artifacts back. All models share the same schema so your code doesn't break when you swap models.
02TRACEABILITY
REVISION HEADERS EVERYWHERE
Every response includes model version, weights checksum, and timestamp. Reproducible for audits and papers.
03ACCESS
STABLE API
Request schema is versioned. Adding a model doesn't break your code. Swap models by changing one field.
GET STARTED
Request API access. We'll send you keys and documentation.