Under Active Development

Protein AI Platform

Unified API for protein language models (ESM3) and structure prediction (AlphaFold2). Run locally on your hardware or scale to cloud GPUs — with built-in LLM integration and interactive 3D visualization.

Local & Cloud Inference Multi-GPU Support Zero-Trust Networking

How it connects

A single gateway routes requests to the right model backend. LLMs connect through an MCP server — a standard protocol that lets AI assistants call your tools directly — over a private VPN. GPU resources auto-stop when idle.

Built for protein research

Multi-Model Gateway

A single API routes to ESM3, AlphaFold2, and ESMFold with automatic health checks and failover. Submit long-running predictions to an async job queue with progress tracking, or run fast queries synchronously.

LLM & Visualization Tools

All models are exposed as conversational tools that any AI assistant can call directly. Predict structures, compare variants, and generate interactive 3D views, per-residue confidence plots, and sequence alignments — all through natural language.

01

Local & Cloud

Run models on your own hardware (CPU or GPU) with no cloud dependency, or scale to cloud endpoints that auto-stop when idle.

02

Multi-GPU

All model services support NVIDIA (CUDA) and AMD (ROCm) GPUs, with automatic CPU fallback. One config switch.

03

Private by Design

All infrastructure runs on private networks behind a zero-trust VPN. No public endpoints. Your sequences stay on your machines.

04

Cross-Platform

Automated installers for Windows and Linux. One-command setup handles dependencies, GPU drivers, and deployment.

Three engines, one API

Each model serves a different stage of the protein design workflow, from rapid screening to high-accuracy structure prediction. All run locally or in the cloud.

ESM3 Protein Design
Meta's protein language model. Generate novel sequences, compute learned representations, and score mutant fitness. Runs locally on CPU or GPU with batched scoring.
generate embed score batch
AlphaFold2 Structure Prediction
High-accuracy 3D structure prediction with per-residue confidence scoring (pLDDT) and batch processing. Runs locally via ColabFold, which handles sequence alignment online — no multi-terabyte databases to download.
predict batch analyze visualize
ESMFold Fast Folding
Single-sequence protein folding in seconds on GPU. Ideal for large-scale screening before committing to a full AlphaFold2 prediction run.
quick fold screen batch compare

Talk to your models

Connect any AI assistant and work with protein models through natural language. Submit prediction jobs, track progress, generate 3D structure views, and compare variants — all conversationally.

Works with Claude, GPT-4, and any LLM that supports the Model Context Protocol (MCP) — the open standard for giving AI assistants access to external tools.

protein-ai — mcp session

> Predict the structure of MKFLILLFNILCLFPVLAADNH and show me the result


submitting prediction job...

Job af-22a7b3 submitted

Status: running (12/22 residues processed)


job complete — retrieving results...

Avg. confidence (pLDDT): 87.4 — high

22 residues, 2546.1 Da


generating 3D visualization...

Interactive viewer ready (confidence coloring)

Confidence plot saved

Built for real research

Protein AI Platform is developed and maintained by Eugenio La Cava. The codebase is tested with 200+ automated tests spanning unit, integration, chaos, and property-based testing. The platform is designed for production use in academic and commercial protein research workflows.

3
Model backends
200+
Automated tests
2
GPU platforms
11
MCP tools

Free for research

Academic and non-profit teams get full access at no cost. Commercial licensing is available for companies building on the platform.

Commercial

Pharma, biotech R&D, SaaS deployment

Custom per org
  • Full platform access
  • All models and tools
  • SaaS deployment rights
  • Priority support available
Contact for Pricing

Citation

La Cava, E. (2026). Protein AI Platform: Unified API for ESM3 and AlphaFold protein prediction services. https://eugenio.github.io/protein-ai-platform