Fastfold AI Docs - Quick Start

Install the FastFold Agent CLI or Python SDK, authenticate, and run your first fold.

The FastFold platform provides a simple, production-ready API to fold protein sequences, compute structural properties, run complex workflows, and orchestrate agents.

FastFold Agent CLI

The fastest way to get started. Ask questions in natural language — the agent plans the analysis, calls the right tools, and returns data-backed conclusions.

Install

uv tool install "fastfold-agent-cli[all]" --python 3.10

Authenticate

fastfold setup

This walks you through setting ANTHROPIC_API_KEY and FASTFOLD_API_KEY interactively.

Run your first query

fastfold "Fold this sequence with boltz-2 and show me the binding pockets: MALWMRLLPLL..."

# Or start an interactive session
fastfold

Full command reference, tools, and interactive mode → Agent CLI

Python SDK and CLI

For scripting, notebooks, and CI pipelines.

Before you can use the API, you need to create an API key. You can do this in the cloud dashboard. Once you have created an API key, you can use to authenticate requests to the API. Make sure to keep your API key secret and do not share it with anyone. You can store in your environment variables (.zshrc, .bashrc, .env, etc.) to use in your code.

export FASTFOLD_API_KEY="your_api_key_here"

The Python SDK will automatically pick up the API key from FASTFOLD_API_KEY by default.

Install the Python SDK

Requires Python 3.8+.

# Recommended (from PyPI)
pip install fastfold-ai

# Local development (from project root)
pip install -e .

Fold a protein sequence

With the SDK installed, create a file named example.py and run:

from fastfold import Client

client = Client()

myJob = client.fold.create(
    sequence="LLGDFFRKSKEKIGKEFKRIVQRIKDFLRNLVPRTES",
    model="boltz-2",
    is_public=True,
)

results = client.jobs.wait_for_completion(myJob.id, poll_interval=5.0, timeout=900.0)

print("Status:", results.job.status)
print("CIF URL:", results.cif_url())
print("Mean PLDDT:", results.metrics().mean_PLDDT)

link = results.get_viewer_link()
print("Open in viewer:", link)

Run the same fold from the CLI

fastfold-cli fold --sequence "LLGDFFRKSKEKIGKEFKRIVQRIKDFLRNLVPRTES" --model boltz-2

On success, the CLI prints the created job ID to stdout.

Next steps

Agent CLI — natural language research workflows: /agent-cli
Python SDK — auth, config, patterns: /python-sdk
Reference examples — Boltz-2 affinity, pockets, indexing, artifacts, metrics: /python-examples
CLI flags and scripting: /cli
API Reference: /api
Available models: cloud.fastfold.ai/models

Quick Start