SCOPe 2.08: Structural Classification of Proteins — extended

Programmatic access: REST API and MCP endpoints

This page lists what can be retrieved from each endpoint of the SCOPe REST API and the tools offered by the public MCP service. For step-by-step instructions on connecting an AI research assistant or installing the research skills, see Use SCOPe with a research assistant.

• Conventions shared by all endpoints
• Example requests
• REST API endpoints: releases, classification hierarchy, domains, sequences, PDB entries, PDB chains, AlphaFold models, ASTRAL, quality and downloads, search
• MCP service and tools
• Access and rate limits

Conventions shared by all endpoints

• Base URL: https://scop.berkeley.edu/api/v1. All paths below are relative to it. The API is read-only and every operation uses HTTP GET. No account or API key is needed.
• The machine-readable OpenAPI 3.1 specification (openapi.yaml) gives the full parameter and response schemas for every endpoint listed here.
• Releases. Every biological data path takes a {release} segment (for example /releases/2.08/sids/d1bz1a_); the sequence endpoints take the same value as an optional release query parameter instead, defaulting to the current release. Any public SCOP 1.x or SCOPe 2.x version is accepted, such as 1.75 or 2.08. For a release that receives periodic updates, such as 2.08, the current periodic update is served. The dated label of the current update (for example 2.08-YYYY-MM-DD, as reported by /releases/current) is also accepted, but a superseded periodic label returns HTTP 404 rather than older data, and stable-only views (2.08-stable) are not yet available. Record the release and retrieval date when citing results.
• Responses are JSON, except for the OpenAPI document (YAML), coordinate downloads (PDB format) and sequences requested with format=fasta. Errors are returned as JSON objects with status, title and detail fields. Records include api_url links to related endpoints.
• Pagination. Collections accept limit and return has_more and an opaque next_cursor; pass that value as cursor to fetch the next page.
• Identifiers. sunid: numeric node identifier (e.g. 46456); sid: domain identifier (e.g. d1bz1a_); sccs: classification string (e.g. a.1.1.2); PDB code (e.g. 1ake); chain: PDB code plus one case-sensitive chain character (e.g. 1akeA); AlphaFold model (e.g. AF-Q9BSH3-F1). See stable identifiers.
• Sequence and structure matches to AlphaFold models and unclassified chains are evidence for investigation; they are not official SCOPe classifications.

Example requests

curl 'https://scop.berkeley.edu/api/v1/releases/current'
curl 'https://scop.berkeley.edu/api/v1/releases/2.08/sids/d1bz1a_'
curl 'https://scop.berkeley.edu/api/v1/releases/2.08/alphafolds/AF-Q9BSH3-F1/hits?limit=5'

These return, respectively, the current release and its latest periodic update; the classification and lineage of domain d1bz1a_ (hemoglobin alpha chain) in SCOPe 2.08; and the top five sequence and structure matches from the human Nicolin-1 AlphaFold model to SCOPe domains.

REST API endpoints

Service metadata and releases

Classification hierarchy (by sunid or sccs)

Domains (by sid)

Sequences

PDB entries

PDB chains and chain comparisons

Chain results describe the requested chain, including chains not yet classified in SCOPe. A cluster_representative link (90% sequence identity) is supplied separately and is never substituted automatically; see Explore PDB chains.

AlphaFold model comparisons

Comparisons of human AlphaFold models with classified SCOPe domains; see Explore human AlphaFold models.

ASTRAL, quality and downloads

These endpoints describe static files; use Downloads for bulk data.

Search

Search is for discovery; fetch the exact release-scoped records above before interpreting a result.

MCP service and tools

The public Model Context Protocol (MCP) service at

https://scop.berkeley.edu/mcp

(Streamable HTTP, no authentication) is a thin layer over the REST API: each tool calls the endpoints above, pins every call to one release, and adds source links and retrieval metadata. It returns the same underlying data, sometimes in a more compact form. Most tools take an optional release argument. Connection instructions are in Connect a web assistant.

Access and rate limits

• The REST API and the MCP service are not behind the browser challenge used for some interactive web pages, so scripts and agents can call them directly.
• The MCP service accepts up to 60 requests per 10 seconds from each IP address; further requests receive HTTP 429.
• The REST API processes up to 20 requests at once; further requests wait up to 5 seconds and then receive HTTP 503.
• Across the whole site, each IP address may hold at most 20 simultaneous connections.
• For whole-release analyses, use the downloadable files rather than paging through the API.

Report problems to scope@compbio.berkeley.edu with the failing request.