---
title: AlphaFold quickstart
description: Fetch a pre-computed AlphaFold structure by UniProt ID in 2-5 seconds, no GPU — pLDDT-colored canvas card, downloadable PDB, ESMFold fallback on a miss.
---

You want the 3D structure of a known protein and it is already in the AlphaFold Database — so instead of predicting it, you [fetch it](/docs/alphafold) by UniProt ID (or protein name) in one step, and the pre-computed structure lands as a canvas card in 2-5 seconds with no GPU.

## Fetch an AlphaFold structure

The fastest way to get a protein structure on the platform when the protein is in the AlphaFold Database.

## In chat (recommended)

Just ask:

> Fetch the AlphaFold structure for **P00533**

Or by protein name (the platform resolves to UniProt automatically):

> Get the AlphaFold structure for human EGFR

The result lands as a canvas card with the 3D structure, pLDDT confidence coloring, and a download button for the PDB file.

## Chat is the only way in

An AlphaFold Database fetch runs only inside a chat turn: there is no tool form, no job, and no CLI or MCP command for it. The one thing Ivy needs is a UniProt accession, or a protein name it can resolve to one.

Typical duration: **2-5 seconds** for the HTTP round-trip + structure parsing. No GPU.

## What happens if the protein isn't in the database

The fetch returns a "not found" result. To handle this gracefully, either:

1. Chain in chat: *"Try AlphaFold first; if it's not in the database, run ESMFold."*
2. Or ask Ivy to [predict it with ESMFold](/docs/esmfold) from the UniProt sequence.

The AlphaFold DB covers ~200M proteins across most well-studied proteomes, so misses are rare for human / model-organism work — but common for novel synthetic sequences or fragments.

## Next steps

After fetching a structure:

- **Dock a ligand**: *"Now dock aspirin against this structure"* → runs [DiffDock](/docs/diffdock)
- **Find binding pockets**: *"Find binding pockets"* → runs [P2Rank](/docs/p2rank)
- **Refine for docking**: *"Add hydrogens and assign partial charges"* → runs [PDB2PQR](/docs/pdb2pqr)
- **Compare to experimental structure**: *"Is there a PDB structure for this protein?"* → runs a [PDB search](/docs/pdb)

## Read next