Skip to main content
The Biohub Platform API gives you programmatic access to hosted protein models, asynchronous batch jobs, and the ESM Atlas. Every API uses the base URL https://biohub.ai.

Get an API key

The Protein Models API and the Batch API authenticate requests with an API key. The documentation uses the terms API key and API token interchangeably.
1

Open the Developer Console

2

Sign in or create an account

Sign in with your Biohub account. If you don’t have an account, create one.
3

Create a key

Under API Keys, create a new key. Copy the key and store it somewhere secure, such as the BIOHUB_API_KEY environment variable. Don’t paste the key into shared code or chats.
Send the key in an Authorization: Bearer header with each HTTP request:

Use the esm Python package

The esm package on PyPI is a Python SDK for the Biohub Platform API. You can call the API directly over HTTP, but the package wraps the endpoints in model clients and adds libraries for working with protein sequences and structures. The package also requests compact pickle response formats automatically.
Connect a client to the model you want to use and pass your API key as the token:
For ESMFold2, use SequenceStructureForgeInferenceClient. For ESM3, use ESM3ForgeInferenceClient. See Get started to set up each model and the Tutorials for examples.

Available APIs

Protein Models

Run on-demand inference with ESMC, ESMFold2, and ESM3, including folding, logits and embeddings, encoding, decoding, generation, and inverse folding.

Batch API

Submit large or long-running jobs asynchronously, such as folding inputs up to 4096 tokens, then track progress and retrieve results.

ESM Atlas

Query predicted protein structures, clusters, sparse autoencoder (SAE) features, and similarity search through public REST endpoints that don’t require an API key.