AlphaFold DB API

Free AlphaFold Protein Structure Database API with no key. Fetch predicted 3D structures, per-residue confidence and PDB/mmCIF download links by UniProt accession. Tested example included.

No API key requiredCORS enabledHTTPSFree tier

Endpoint tested and returned HTTP 200 on 2026-08-21

What is the AlphaFold DB API?

The AlphaFold Protein Structure Database API is a free, key-free REST endpoint from EMBL-EBI and Google DeepMind. Given a UniProt accession it returns the predicted structure entry for that protein, including confidence metrics and direct download URLs for the PDB, mmCIF and per-residue error files.

AlphaFold DB holds predicted structures for over 200 million proteins, which is roughly every sequence in UniProt. That coverage is the point: experimental structure determination has solved a few hundred thousand proteins in sixty years, so for most proteins a prediction is the only three-dimensional information that exists. The API is a thin lookup layer over that archive — you hand it a UniProt accession and it hands back a record describing the model plus the URLs of the actual coordinate files.

Everything here is a prediction, and the API is unusually good about saying so. `globalMetricValue` is the mean pLDDT, a self-reported per-residue confidence on a 0-100 scale, and the four `fractionPlddt*` fields tell you what proportion of the chain falls into each confidence band. The example below returns 67.38 mean pLDDT with 35.7% of residues in the very-low band, which is a signal that large stretches of that protein are intrinsically disordered rather than badly modelled. Treat regions below 70 as unreliable and regions below 50 as effectively a ribbon drawn through empty space.

Quick facts

Base URL: https://alphafold.ebi.ac.uk/api
Authentication: No API key or account. The database is a joint EMBL-EBI and Google DeepMind resource released under CC BY 4.0, so you may reuse the models commercially with attribution.
Rate limit: No published hard limit. EMBL-EBI asks automated clients to identify themselves with a contact address in the User-Agent and to download the bulk archives rather than looping over accessions for whole-proteome work.
Pricing: Free. Structures and metadata are licensed CC BY 4.0.
CORS: Enabled — callable directly from browser JavaScript
Official docs: Read the docs

How to use the AlphaFold DB API

Every request below was executed against the live API on 2026-08-21, and the response shown is the real body it returned — not an illustration.

1. Fetch the predicted structure entry for human amyloid-beta precursor protein

GET https://alphafold.ebi.ac.uk/api/prediction/P05067

curl

curl 'https://alphafold.ebi.ac.uk/api/prediction/P05067'

JavaScript (fetch)

const res = await fetch("https://alphafold.ebi.ac.uk/api/prediction/P05067");
if (!res.ok) throw new Error(`Request failed: ${res.status}`);
const data = await res.json();
console.log(data);

Python (requests)

import requests

res = requests.get("https://alphafold.ebi.ac.uk/api/prediction/P05067", timeout=20)
res.raise_for_status()
print(res.json())

Response — HTTP 200 (truncated)

[
  {
    "toolUsed": "AlphaFold Monomer v2.0 pipeline",
    "providerId": "GDM",
    "entityType": "protein",
    "isUniProt": true,
    "modelEntityId": "AF-P05067-F1",
    "modelCreatedDate": "2025-08-01T00:00:00Z",
    "sequenceVersionDate": "1991-11-01T00:00:00Z",
    "globalMetricValue": 67.38,
    "fractionPlddtVeryLow": 0.357,
    "fractionPlddtLow": 0.088,
    "fractionPlddtConfident": 0.266,
    "fractionPlddtVeryHigh": 0.288,
    "latestVersion": 6,
    "allVersions": [
      1,
      2,
      3,
      4,
      5,
      6
    ],
    "sequence": "MLPGLALLLLAAWTARALEVPTDGNAGLLAEPQIAMFCGRLNMHMNVQNGKWDSDPSGTKTCIDTKEGILQYCQEVYPELQITNVVEANQPVTIQNWCKRGRKQCKTHPHFVIPYRCLVGEFVSDALLVPDKCKFLHQERMDVCETHLHWHTVAKETCSEKSTNLHDYGMLLPCGIDKFRGVEFVCCPLAEESDNVDSADAEEDDSDVWWGGADTDYADGSEDKVVEVAEEEEVAEVEEEEADDDEDDEDGDEVEEEAEEPYEEATERTTSIATTTTTTTESVEEVVREVCSEQAETGPCRAMISRWYFDVTEGKCAPFFYGGCGGNRNNFDTEEYCMAVCGSAMSQSLLKTTQEPLARDPVKLPTTAASTPDAVDKYLETPGDENEHAHFQKAKERLEAKHRERMSQVMREWEEAERQAKNLPKADKKAVIQHFQEKVESLEQEAANERQQLVETHMARVEAMLNDRRRLALENYITALQAVPPRPRHVFNMLKKYVRAEQKDRQHTLKHFEHVRMVDPKKAAQIRSQVMTHLRVIYERMNQSLSLLYNVPAVAEEIQDEVDELLQKEQNYSDDVLANMISEPRISYGNDALMPSLTETKTTVELLPVNGEFSLDDLQPWHSFGADSVPANTENEVEPVDARPAADRGLTTRPGSGLTNIKTEEISEVKMDAEFRHDSGYEVHHQKLVFFAEDVGSNKGAIIGLMVGGVVIATVIVITLVMLKKKQYTSIHHGVVEVDAAVTPEERHLSKMQQNGYENPTYKFFEQMQN",
    "sequenceStart": 1,
    "sequenceEnd": 770,
    "sequenceChecksum": "7dd43312cd29a262acdc0517230bc5ca",
    "isUniProtReviewed": true,
    "gene": "APP",
    "uniprotAccession": "P05067",
    "uniprotId": "A4_HUMAN",
    "uniprotDescription": "Amyloid-beta pre

Parameters

Parameter	Type	Required	Description
`qualifier`	path segment	Required	A UniProt accession. The API resolves it to the current model version for that sequence. `P05067`
`sequence_checksum`	query	Optional	CRC64 checksum of a sequence, as an alternative to looking a model up by accession. `7dd43312cd29a262`

Response fields

modelEntityIdstring: The AlphaFold entry identifier, formed as AF-{accession}-F{fragment}. Very long proteins are split into overlapping fragments.
globalMetricValuefloat: Mean pLDDT across the chain, 0-100. Above 90 is a confident backbone and side-chain placement; below 50 usually means disorder rather than a modelling failure.
fractionPlddtVeryLow / Low / Confident / VeryHighfloat: Proportion of residues in each pLDDT band. Read these before trusting any single region of the model.
pdbUrl / cifUrl / bcifUrlstring: Direct download links for the coordinates in PDB, mmCIF and binary CIF form.
paeDocUrlstring: Predicted aligned error matrix, which tells you how confident the model is about the relative position of two domains — the metric that matters for multi-domain proteins.
latestVersion / allVersionsinteger / array: Model version. AlphaFold entries are re-predicted as the pipeline improves, so pin a version if you need reproducibility.

What you can build with the AlphaFold DB API

Show a predicted 3D structure next to a protein record in a bioinformatics tool
Bulk-fetch coordinate files for a shortlist of drug targets
Filter a proteome down to proteins with confidently modelled domains before docking
Flag likely intrinsically disordered regions from the pLDDT fraction fields
Cross-reference a UniProt accession to a downloadable structure without scraping the website

Common errors and how to fix them

404

No model exists for that accession.

Fix: Coverage follows UniProt reference proteomes; obscure TrEMBL entries, viral polyproteins and very long sequences may have no model. Check the accession resolves in UniProt first.

422

The path qualifier is not a recognised accession format.

Fix: Send the bare accession such as P05067, with no version suffix and no `UNIPROT:` prefix.

Empty array

The request succeeded but returned `[]`.

Fix: The API answers with a JSON array, so an existing-but-unmodelled accession can produce an empty list rather than a 404. Check length before indexing.

Slow or truncated downloads

The coordinate files are fetched from a separate static host.

Fix: The URLs in `pdbUrl` and `cifUrl` point at a CDN, not this API. Retry those separately rather than re-requesting the metadata.

AlphaFold DB API — frequently asked questions

Is the AlphaFold API free to use?

Yes. There is no key, no account and no quota to apply for, and the models are released under CC BY 4.0 so commercial reuse is permitted with attribution. EMBL-EBI only asks that heavy automated users identify themselves in the User-Agent.

What does the pLDDT confidence score mean?

pLDDT is the model's own estimate of how well it has placed each residue, on a 0-100 scale. Above 90 is confident down to side-chain orientation, 70-90 is a reliable backbone, 50-70 is low confidence, and below 50 usually indicates an intrinsically disordered region rather than an error.

Can I use an AlphaFold model instead of an experimental structure?

It depends entirely on what you are doing. Confidently predicted domains are often close to crystal structures, but predictions do not capture ligands, cofactors, alternative conformations or the effect of mutations. For anything where the answer matters, check whether an experimental structure exists in the PDB first.

How do I get the actual 3D coordinates?

The metadata record contains `pdbUrl`, `cifUrl` and `bcifUrl` fields pointing at the coordinate files on EBI's static host. Fetch those URLs directly — the API itself never returns atomic coordinates inline.

Tools that pair with this API

JSON Formatter

Format, beautify and minify JSON online with 2-space, 4-space or tab indentation. Sort keys alphabetically and catch syntax errors instantly — free and private.

JSON to CSV Converter

Convert a JSON array of objects to CSV online. Automatic column headers from the union of all keys, delimiter choice and proper quoting — all in-browser.

DNA to RNA Converter

Transcribe DNA to mRNA, get the complement and reverse complement, translate codons to protein and see GC content — all in your browser.

AlphaFold DB is an independent third-party service and is not affiliated with ByteTools or ByteVancer. Details on this page were verified on 2026-08-21; always check the official documentation before relying on this API in production, as terms and limits can change.