What it does
onet-match maps unstructured text about work, including job titles, resume bullets, job postings, project descriptions, interest statements and skill lists, to standardized O*NET-SOC occupation codes across 923 occupations. A fine-tuned bi-encoder retrieves the 20 nearest occupations, and a fine-tuned cross-encoder reranks them. Each match includes O*NET tasks, skills, knowledge, RIASEC interests and a calibrated confidence score.
It suits career exploration in education, resume and profile parsing, skills-based hiring, job-posting normalization, batch coding of large text collections for labor-market analysis, and grounding LLM agents in O*NET data. In the publisher's evaluation, the served model ranked the correct occupation first in 51.1% of held-out O*NET text cases, compared with 20.5% for a TF-IDF baseline, and in 65.4% of messy, human-style inputs, compared with 17.4%. It takes about 47 ms per query on a laptop.
It can be used from Python, a command-line tool, a REST API, Docker, or an MCP server with two read-only tools, match_occupations and get_occupation. The project runs locally and needs no API key or account to match. The first call downloads a pretrained model bundle and caches it.
Limitations and requirements
The first call downloads a pretrained model bundle of about 0.5 GB from the Hugging Face Hub and caches it locally.
Safety
Our team has not reviewed this resource. Only automated rules checked it; resources with high-risk signals are not published. These are signals, not a guarantee that installing it is safe. Try it in a sandbox project first.
- Runs code or changes an assistant's behaviour
Sources
- Official MCP Registry first seen 13 Nov 2026 · last seen 13 Nov 2026 · its page on the source
onet-match
Occupation classifier: free text to ranked O*NET-SOC codes with tasks, skills and RIASEC. Local.
v0.3.5As shown on registry.modelcontextprotocol.io when we last read it