Introduction

The Boltz API lets you predict structures, affinities and other properties of biomolecular complexes, as well as design and screen small molecules and proteins.

Quickstart

Let's make our first prediction with the Boltz API. This should take around 60 seconds. First, install the CLI:

macOS / Linux

curl -fsSL https://install.boltz.bio/boltz-api/install.sh | sh
boltz-api auth login

Windows

irm https://install.boltz.bio/boltz-api/install.ps1 | iex
boltz-api auth login

After signing up for an account in the API Console, you'll be granted credits so you can test the API for free. Then:

macOS / Linux

boltz-api predictions:structure-and-binding run \
    --model boltz-2.1 --name my-prediction \
    --input @yaml://<(cat <<EOF
entities:
  - type: protein
    value: MKTIIALSYIFCLVFA
    chain_ids: ["A"]
  - type: ligand_smiles
    value: CC(=O)OC1=CC=CC=C1C(=O)O
    chain_ids: ["B"]
EOF
)

Windows

$yaml = @"
entities:
  - type: protein
    value: MKTIIALSYIFCLVFA
    chain_ids: ["A"]
  - type: ligand_smiles
    value: CC(=O)OC1=CC=CC=C1C(=O)O
    chain_ids: ["B"]
"@
boltz-api predictions:structure-and-binding run `
    --model boltz-2.1 --name my-prediction `
    --input $yaml

Once the prediction is finished, explore the local files:

tree boltz-experiments/my-prediction

Output

boltz-experiments/my-prediction
├── outputs
│   ├── archive.tar.gz
│   └── files
│       └── prediction
│           ├── metrics.json
│           ├── sample_0_pae.npz
│           └── sample_0_predicted_structure.cif
└── run.json

4 directories, 5 files

Choose your interface

Our API guides all follow the same pattern: choose Python, CLI, or TypeScript at the top of the page, and the examples follow your choice.

• Python SDK — submit a job, poll and download results with a high level SDK (run(), or start() then client.experiments.download_results()); or drive the API directly with retrieve() and list_results(). Best for notebooks and scripts.
• CLI (boltz-api) — write your input as a YAML file and pass its path; download-results polls, downloads, and resumes on your behalf. Best for the terminal and shell pipelines, including coding agents.
• TypeScript SDK — a direct REST-style client: start(), retrieve(), and listResults(). Best for apps and backend services.

Run a job

• Predict structure and binding — Submit a biomolecular complex and get back predicted structures and binding affinities.
  Predict structure and binding →
• Design novel molecules and proteins — Generate new small molecules or protein binders against your targets of interest using our generative models.
  Small molecule design → · Protein design →
• Screen a library — Score small molecules or protein sequences against a given target.
  Small molecule screen → · Protein screen →
• Predict ADME — Score a batch of small molecules for Tier-1 ADME summary properties.
  Small molecule ADME →

Next steps

• Authentication — API key types and SDK setup.
• Agent integrations — Use Boltz from Claude Code, Codex, and other coding agents.
• API Reference — Full endpoint documentation with SDK examples.