worldoftaxonomy

name	worldoftaxonomy
version	1.0.0
description	Query 1,000+ global classification systems (NAICS, ISIC, NACE, HS, ICD, SOC, ISCO, CPC, UNSPSC, and more) with 1.3M+ nodes and 326K+ crosswalk edges. Classify businesses, products, occupations, diseases, or documents under standard codes. Translate codes across country and system boundaries. Every one of the 434 domain taxonomies is bridged to NAICS/ISIC/NACE via sector anchors, and every edge carries an `edge_kind` label.
author	Ram Katamaraja (Colaberry AI)
homepage	https://worldoftaxonomy.com
license	MIT

WorldOfTaxonomy Skill

What this does

Turns any free-text question about industry, occupation, medical, trade, or regulatory classification into a grounded answer backed by 1,000+ official taxonomies linked as a single knowledge graph.

When to invoke

Use this skill when the user wants to:

• Assign a standard code to a business description, product, occupation, diagnosis, or document
• Translate a code between systems (NAICS 2022 <-> ISIC Rev 4, SOC 2018 <-> ISCO-08, ICD-10-CM <-> ICD-10-GM, HS 2022 <-> CPC v2.1, etc.)
• Find every cross-system equivalent of a given code
• Browse hierarchy (children, ancestors, siblings, subtree) of a system
• Compare two systems' coverage or structure
• Audit data provenance and crosswalk density across systems

Do NOT invoke for general explanations of what NAICS/ISIC/ICD are - answer from general knowledge unless the user wants live data.

How to use

Primary: REST API

Base URL: https://worldoftaxonomy.com/api/v1

Auth: Authorization: Bearer wot_<32hex> (get a key at https://worldoftaxonomy.com/dashboard). Anonymous access works at 30 req/min; authenticated keys get 1000-50000 req/min depending on tier.

Core endpoints:

GET  /systems                                       # list systems
GET  /systems/{id}                                  # detail + roots
GET  /systems/{id}/nodes/{code}                     # node detail
GET  /systems/{id}/nodes/{code}/children
GET  /systems/{id}/nodes/{code}/ancestors
GET  /systems/{id}/nodes/{code}/siblings
GET  /systems/{id}/nodes/{code}/equivalences        # crosswalk edges
GET  /search?q=<term>&system=<optional>
POST /classify                                      # body: {"description": "..."}
GET  /equivalences/stats                            # coverage matrix

OpenAPI spec: https://worldoftaxonomy.com/api/v1/openapi.json

Secondary: MCP server

For MCP-aware clients, install world_of_taxonomy from the GitHub repo and run python -m world_of_taxonomy mcp as a stdio server. Tools include classify_business, translate_code, translate_across_all_systems, get_equivalences, get_crosswalk_coverage, get_system_diff, explore_industry_tree, and list_crosswalks_by_kind.

Response guidance

• Always cite the system ID and code (e.g. "NAICS 2022 -> 541511") and the source/authority where relevant.
• When translating a code, list all equivalents returned by /equivalences - don't pick just one unless the user asks for the best match.
• When classifying from free text, return the top 3-5 candidates with confidence ranking.
• Flag provenance tier when it matters: official_download > structural_derivation > manual_transcription > expert_curated.

Two-category model: domain taxonomies vs official standards

Every system carries a category field: "domain" or "standard".

• "domain" systems (IDs start with domain_) are plain-language on-ramps curated by WorldOfTaxonomy: short, written in industry vocabulary, crosswalked into the relevant NAICS/ISIC/etc.
• "standard" systems are the canonical codes published by governments and standards bodies (NAICS, ISIC, NACE, SIC, SOC, ICD, HS, CPC, and peers).

POST /classify and the MCP classify_business tool return domain_matches and standard_matches as separate arrays (no flat matches field). Surface domain matches first in the user's reply so they can recognize their business, then show the standard codes they will need to report to auditors or agencies. Full reference: https://worldoftaxonomy.com/guide/domain-vs-standard.

Four edge kinds

Every equivalence response carries an edge_kind computed from both endpoint systems:

• standard_standard - classic statistical crosswalk (NAICS <-> ISIC, ISIC <-> NACE, HS <-> CPC, SOC <-> ISCO, ...)
• standard_domain - bridge from an official code to a curated domain taxonomy
• domain_standard - reverse of the above
• domain_domain - reserved for future cross-domain edges; none generated yet

Filter any equivalence or translation endpoint with ?edge_kind=standard_standard (or a comma-separated list). Use the MCP tool list_crosswalks_by_kind to count + sample a specific kind, optionally narrowed to a system. Generated domain-bridge edges are stamped match_type='broad' with provenance derived:sector_anchor:v1 (direct) or derived:sector_anchor:v1:fanout (ISIC/NACE parallels). Filter ?match_type=exact to exclude every generated bridge.

Repository

Source, data, and issue tracker: https://github.com/colaberry/WorldOfTaxonomy