boltz

Predict biomolecular structures with Boltz-1 and Boltz-2 on local GPUs.

11|Updated Mar 4, 2026
One-click install
npx skills add https://github.com/junior1p/ProteinClaw --skill boltz-junior1p
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: boltz
Source: https://github.com/junior1p/ProteinClaw/tree/main/skills/boltz
Command: npx skills add https://github.com/junior1p/ProteinClaw --skill boltz-junior1p

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Boltz Structure Prediction provides local, open-source protein structure prediction to accelerate design validation and reduce reliance on external services.

Core Features & Use Cases

  • Predict protein and complex structures with Boltz-1 and Boltz-2.
  • Validate designed binders and explore structural hypotheses on GPU-enabled hardware.
  • Use open-source, offline workflows as an alternative to cloud-based predictors.

Quick Start

Install Boltz and run boltz predict on your local FASTA file to generate structure predictions.

Frequently Asked Questions about boltz

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run protein structure prediction locally on a GPU?

Run local protein structure prediction by executing the boltz predict command on a FASTA file. This requires a CUDA-enabled GPU (CUDA 12.x) and Python 3.10+ to generate CIF models, confidence.json, and pae.npy outputs.

Can I validate designed protein binders offline without cloud services?

Yes, you can validate designed protein binders offline using open-source, GPU-accelerated structure prediction. This local workflow serves as an alternative to cloud-based predictors, reducing reliance on external services.

What hardware do I need for biomolecular structure prediction with Boltz?

Biomolecular structure prediction with Boltz requires a local environment with CUDA-enabled GPUs (CUDA 12.x) and sufficient VRAM. You also need Python 3.10 or higher installed to execute the predictions.

What output files are generated when predicting protein complex structures?

Predicting protein complex structures generates CIF models, confidence.json, and pae.npy files. These outputs provide the structural coordinates and confidence metrics needed for analysis.

Is there an open-source alternative to cloud-based protein structure predictors?

Yes, Boltz offers an open-source alternative to cloud-based protein structure predictors. It runs locally on your hardware, using Boltz-1 and Boltz-2 models for offline biomolecular structure prediction.

Why does local structure prediction require significant VRAM?

Local structure prediction requires significant VRAM because the GPU-accelerated Boltz-1 and Boltz-2 models process complex biomolecular computations. Insufficient VRAM will prevent the local environment from running predictions effectively.