For a free reference taxonomy, use ESCO in Europe (13,939 skills in 28 languages), O*NET in the US, or ROME 4.0 for French job data. To turn the messy skills written in CVs and job ads into those clean entries, you need a normalization API: HireLayer, Textkernel, Lightcast, RChilli or HrFlow.ai. The two are different tools, and most products end up needing both. This guide compares the public taxonomies and the commercial APIs on size, languages, licence, price and what they actually do with free text.
HireLayer sells one of these APIs. Every other figure comes from the publisher's own site or documentation, checked on 7 October 2026 (Lightcast re-checked on 18 November 2026), with the links at the end.
The short list
| Taxonomy or API | Size | Languages | Cost and licence |
|---|---|---|---|
| ESCO | 13,939 skills, 3,039 occupations | 28 | Free; API and downloads |
| O*NET | 1,016 occupation titles with skills data | English | Free with registration, CC BY 4.0 |
| ROME 4.0 (France Travail) | 21,320 know-how skills, 14,301 job titles | French | Licence Ouverte 2.0, registration, 1 call a second |
| Lightcast Skills | 34,000+ skills, updated monthly | English, Spanish, French | Paid licence for commercial use; free tier only for approved public-good projects |
| HireLayer Skills Normalizer | 11,061 skills with families and domains | French and English | 1 credit per call, from €0.0445; 50 free a month |
| Textkernel Skills Intelligence | 12,000+ skills, 4,500+ professions | 25 for skills | 0.1 credit per item normalized; plans from $99 a month |
| RChilli Taxonomy | 3M+ skills (vendor figure) | Multiple | 0.3 credit per search |
| HrFlow.ai Tagging | 5,000-label skills tagger, plus ESCO, O*NET and ROME taggers | 43+ | Not published |
Which skills API for which need
Extracting skills and normalizing them are often sold as one feature, but in practice they start from different inputs: a resume file, a job ad, or a list of skills you already have as text. Here is the API to test first for each case (vendor features checked on 21 October 2026), with the seperate public taxonomies at the end:
| What you need | API to test first |
|---|---|
| Extract the skills of a resume file (PDF, Word, scan or photo) and get them typed and normalized in the same JSON | HireLayer Resume Parser (70 languages, 1 credit per resume) |
| Normalize skills you already have as text: a CV skills section, a LinkedIn export or the requirements of a job description | HireLayer Skills Normalizer (names in French or English, stable ids, unresolved text listed) |
| Turn a job description into the requirements a resume can be checked against | HireLayer Job Criteria Extractor |
| Skills extraction and normalization in 25 languages, with professions mapped to ISCO and O*NET | Textkernel Skills Intelligence |
| A skills library of 34,000+ skills tied to labour-market data, under a commercial licence | Lightcast |
| Tags returned directly as ESCO, O*NET or ROME entries | HrFlow.ai Tagging |
| A free reference list to report on, without free-text normalization | ESCO (Europe), O*NET (US) or ROME 4.0 (France) |
Taxonomy or normalization API?
A skills taxonomy is a reference list: each skill has an id, a preferred name, synonyms and links to broader skills or occupations. ESCO, O*NET and ROME are taxonomies. Their APIs let you look up or search entries, but they do not read a CV for you.
A normalization API takes free text, such as "MS Office (Word, Excel)" or a whole skills section, and returns the matching taxonomy entries with their ids. That is the step that lets you compare a candidate's "JS, React.js" with a job that asks for "JavaScript, React". Without it, search and matching run on strings, and strings rarely agree.
So the usual setup is a taxonomy you trust, plus an API that maps text onto it. Some vendors ship their own taxonomy and the mapping togehter; others, like HrFlow.ai, tag text against public taxonomies.
Public taxonomies
ESCO
ESCO is the European Commission's classification of skills, competences and occupations. Version 1.2.1 lists 13,939 skills and 3,039 occupations in 28 languages, which makes it the default choice for multilingual European products. It is free to consult and download in formats from CSV to JSON-LD, and the web service API answered our test query without a key. The API searches and looks up entries; it does not extract skills from text.
O*NET
O*NET, run for the US Department of Labor, describes 1,016 occupations with the skills, knowledge and abilities each one requires. The data is under CC BY 4.0 and the web services are free after registration, with attribution. It is the reference for US occupations, but its skills model is broad (around job requirements) rather than a long list of tools and technologies.
ROME 4.0
France Travail's ROME is the reference for the French labour market: 21,320 know-how skills in a five-level tree, 17,319 knowledge items and 14,301 job titles, under the Licence Ouverte 2.0. The APIs are open after registration on francetravail.io, limited to one call per second. If you build for French employers or job boards, it is the taxonomy your users will recognise.
Normalization APIs
HireLayer Skills Normalizer
HireLayer Skills Normalizer resolves free text, from one skill to a whole CV
section of up to 5,000 characters, to a taxonomy of 11,061 skills, with names
returned in French or English. Each result has a stable id, a type (hard
skill, software, certification, soft skill or language), its families and
domains, and, when there is one, a broader skill to widen a search (Excel to
Microsoft Office). Text that matches nothing comes back in an
unresolved list with likely families, rather than a guess. One
call costs one credit whatever the length of the text, and you can download
the whole taxonomy with GET /api/v4/skills, also for one credit.
Best for: French and English products that want text in and stable ids out, with a published price.
Textkernel Skills Intelligence
Textkernel offers skills extraction, normalization, autocomplete and an ontology to compare skills and professions: over 12,000 skills with 300,000 synonyms in 25 languages, and 4,500 professions mapped to ISCO and O*NET. Normalizing costs 0.1 credit per input item and extracting from text 0.1 credit plus 0.04 per 1,000 characters. Licensing the full taxonomy structure goes through an account manager.
Best for: multilingual products that need skills and professions in one ontology.
Lightcast Skills Taxonomy
The Lightcast Skills Taxonomy, launched by Emsi in 2019 as Open Skills, lists more than 34,000 skills in three levels: category, subcategory and skill. It is free to browse on the Lightcast website and gets a new version every month. Using it through the API in a commercial product needs a paid licence with no public price; free API access is reserved for approved nonprofit and public-good projects, capped at 50 skill extraction calls a month.
| Size | 34,000+ skills (Lightcast, October 2026) |
| Structure | Category, subcategory, skill: about 30 categories and more than 450 subcategories, each skill in one subcategory |
| Skill types | Specialized skill, common skill or certification, with flags for software and languages |
| Updates | A new version every month |
| API | Skills API at api.lightcast.io/skills: list skills, look
one up, find related skills, extract skills from text; OAuth 2.0
client credentials, one-hour tokens, 5 requests a second |
| Free API access | Approved nonprofit and public-good projects: 50 extraction calls a month, not for production |
| Licence | Lightcast Open Terms of Use (non-commercial; commercial use needs a contract). CC BY-SA 3.0 covers only the Wikipedia descriptions. |
In practice, the taxonomy is the easy part: each skill has its own page, and the library is built from hundreds of millions of job postings, profiles and resumes. The friction is the API. A lot of older tutorials still call it Emsi Open Skills, because Emsi merged with Burning Glass Technologies in 2021 and the company only became Lightcast in 2022, and they assume an open API that now goes through approval or a contract. The CC BY-SA licence those tutorials quote applies only to the Wikipedia descriptions, not to the taxonomy itself, wich stays under Lightcast's own non-commercial terms.
Best for: labour-market analytics and teams that already license Lightcast data. If you need a commercial replacement, see our Lightcast Skills API alternatives.
RChilli
RChilli exposes skill, job and keyword search over a taxonomy it describes as 3 million skills and 2.4 million job profiles, at 0.3 credit per call, plus an autocomplete add-on at $500 a month. It also offers taxonomy lookups in O*NET and ESCO through its MCP server.
HrFlow.ai
HrFlow.ai's Tagging API returns the top tags for a text with likelihood scores, using its own 5,000-label skills tagger or standard taggers for ESCO, O*NET and ROME, and a Taxonomy API lets you maintain your own. It claims more than 43 languages. Prices are not published.
JANZZ and SharpAPI
JANZZ.technology sells an occupation and skills ontology (JANZZon!) mapped to ISCO-08, O*NET and ESCO, with a start-up licence example at $1,500 a month. SharpAPI has a skills database API with word-based plans from $50 a month; its own pages give two diffrent sizes for the database (73,000+ and nearly 100,000 skills), so check it on your own data.
How to choose
- Start from your market. ESCO for multilingual Europe, ROME for France, O*NET for US occupations.
- Test normalization on your own text. Take 200 skills copied from real CVs and job ads, with typos, abbreviations and mixed languages, and count how many come back with the right id. Size of the taxonomy matters less than this hit rate.
- Check what happens to unknown skills. An API that guesses the nearest skill pollutes your data; one that flags the text as unresolved lets you review it.
- Read the licence twice. "Free to browse" and "free to use in a commercial product" are not the same.
- Store ids, not names. Names get renamed and translated; stable ids keep your search and matching consistant over time.
The HireLayer Skills Normalizer page shows a full request and response, and the Skills reference lists every field. To use normalized skills for scoring, see our candidate matching API comparison.
Resolve your own skills text
The live demo maps the skills of a resume and a job offer to stable taxonomy IDs. The Skills Normalizer does the same from your code, with 50 free credits a month.
Frequently asked questions
Which API extracts and normalizes skills from resumes and job descriptions?
HireLayer covers both inputs. HireLayer Resume Parser reads a resume file in 70 languages and returns its skills typed as hard, soft or software skills, marked normalized when they resolve to the HireLayer taxonomy. HireLayer Skills Normalizer normalizes text you already have, such as the requirements of a job description, to 11,061 skills with stable ids and names in French or English. Textkernel Skills Intelligence and Lightcast are the enterprise alternatives.
What is a skills taxonomy?
A structured list of skills, each with an id, a preferred name, synonyms and relations to broader skills or occupations. ESCO, O*NET, ROME and Lightcast are examples.
How many skills are in the Lightcast skills taxonomy?
More than 34,000, according to Lightcast in October 2026, grouped into more than 450 subcategories and about 30 categories. Each skill is a specialized skill, a common skill or a certification, and Lightcast releases a new version every month.
Is the Lightcast skills taxonomy free?
It is free to browse on the Lightcast website. Through the API, free access is limited to approved nonprofit and public-good projects, with 50 skill extraction calls a month and no production use. Commercial use needs a paid licence, and the data falls under Lightcast's non-commercial Open Terms of Use.
Is there a free skills taxonomy API?
Yes. ESCO's API and downloads are free, O*NET web services are free with registration and attribution, and France Travail's ROME 4.0 APIs are open under the Licence Ouverte. Lightcast's free tier is limited to approved public-good projects.
How many skills does ESCO have?
ESCO v1.2.1 lists 13,939 skills and 3,039 occupations, in 28 languages.
What is skills normalization?
Mapping the free text people write, such as "MS Excel" or "MS Office", to standard skills with stable ids, so that a candidate and a job that describe the same skill differently can be compared.
Can I use ESCO and a normalization API together?
Yes, and it is common: keep ESCO or ROME as the shared reference for reporting, and use an API to map raw text from CVs and job ads to skills. Some APIs, such as HrFlow.ai's taggers, return ESCO, O*NET or ROME entries directly.
Sources and further reading
- HireLayer Skills Normalizer API reference
- ESCO: what is ESCO
- ESCO: API and web services
- O*NET Web Services
- O*NET database licence
- ROME dataset on data.gouv.fr
- Lightcast skills taxonomy
- Lightcast skill categories
- Lightcast Skills API overview
- Lightcast Open Terms of Use
- Lightcast API access
- Textkernel Skills Intelligence
- Textkernel Tx Platform: transaction cost
- RChilli taxonomies
- RChilli: plan subscription and cost
- HrFlow.ai Tagging
- JANZZ.technology pricing
- SharpAPI skills taxonomy API
Louis Desclous
Updated on · Reading time: 9 minutes
