Command Palette

Search for a command to run...

UnylyUnyly
Весь каталог

Job Postings

БесплатноНе проверен

Open dataset of 8.4M job postings from employers' own career pages (ODbL 1.0, on Hugging Face), plus a live jobs API and MCP server

GitHubEmbed

Описание

Open dataset of 8.4M job postings from employers' own career pages (ODbL 1.0, on Hugging Face), plus a live jobs API and MCP server

README

Headcount Labs

Headcount Labs Job Postings

An open dataset of 8,367,463 job postings read straight from employers' own career pages and hiring systems, plus a live API and MCP server that serve the same jobs, updated daily.

Star this repository to follow new releases of the dataset.

Guides and comparisons

What Headcount Labs is

Headcount Labs reads the public job boards of employers on Workday, Greenhouse, Lever, iCIMS, SmartRecruiters, Oracle, SAP SuccessFactors and more than 100 other hiring platforms, plus company career sites and government job exchanges. Copies of the same opening on several boards are merged into one job, and each job keeps a link to every original posting. Each field says whether the employer published it or we derived it.

The live API counted 8,365,616 open jobs on 2026-10-19 at 5:24 PM ET (GET /v1/jobs/count).

The dataset

The full dataset lives on Hugging Face, not in this repository. This repository holds the documentation, example queries and a 1,000-row sample.

Table One row per Rows
jobs job, after merging copies 8,367,463
postings listing of a job on one source 9,987,268
job_locations job location (country, region, city, zip, metro area) 9,699,125
companies company ID seen on a job 47,621

Snapshot taken on 2026-10-19 between 12:47 PM and 4:43 PM ET. It is a one-time full snapshot of every job the public API served at that time. Job description text is not included; every posting links to the original listing.

Schema summary

The full column list is in SCHEMA.md. The columns you will use most:

Table Column Meaning
jobs job_id Stable job ID. Join key to postings and job_locations.
jobs title, employer_name Title and employer as the employer published them.
jobs work_arrangement remote, hybrid or in_person when stated, else null.
jobs posted_at Earliest posting date of the job. Files are grouped by its month.
jobs currentness current, uncertain or not_current.
jobs posting_url Link to an original listing.
postings pay_min, pay_max, pay_currency, pay_interval Pay as published, not converted.
postings employment_type, source_type, source_class Employment type as stated, the platform, and the kind of source.
postings application_url Where to apply.
job_locations country, region, locality, zip, cbsa_code ISO country, ISO 3166-2 region such as US-TX, city, postal code, US metro area code.
companies employer_name, job_count Most common employer name for the company and its number of jobs.

Columns ending in _json hold the value with its origin: {"state": "known", "value": ...} when the source published it, derived when we computed it, source_omitted when the source does not publish the field.

Sample

sample/jobs-sample-1000.csv has 1,000 random jobs first posted in October 2026 (flat columns from jobs), so you can see the shape before you download anything.

Example queries

Each query below reads the Parquet files straight from Hugging Face. They were run on 2026-10-19 with DuckDB 1.4.1, pandas 2.3.3 and datasets 5.0.0; the results shown are from that run. The SQL is in examples/queries.sql and the pandas script in examples/pandas_example.py.

Top employers by number of jobs (DuckDB):

SELECT employer_name, job_count
FROM 'hf://datasets/HeadcountLabs/job-postings/data/companies/part-0.parquet'
ORDER BY job_count DESC
LIMIT 10;
employer_name job_count
H&R Block 41,500
Domino's 27,985
The Kroger Co. 26,865
Walgreens 21,434
Marriott 20,904

US jobs by state (DuckDB, all 8 location files, about 25 seconds):

SELECT region, count(DISTINCT job_id) AS jobs
FROM 'hf://datasets/HeadcountLabs/job-postings/data/job_locations/*.parquet'
WHERE country = 'US' AND region IS NOT NULL
GROUP BY region
ORDER BY jobs DESC
LIMIT 10;
region jobs
US-CA 459,517
US-TX 437,948
US-FL 283,916
US-NY 265,725
US-OH 192,004

Posted pay on October 2026 postings, by pay interval (DuckDB, one of the eight October files):

SELECT coalesce(pay_interval, 'no pay published') AS pay_interval, count(*) AS postings,
       round(median(pay_min)) AS median_pay_min
FROM 'hf://datasets/HeadcountLabs/job-postings/data/postings/part-202610-s0-0.parquet'
GROUP BY 1
ORDER BY postings DESC;
pay_interval postings median_pay_min
no pay published 208,753
hour 64,318 19
year 38,798 82,500
month 12,163 39,000
unknown 11,920 49,590

Pay is in the currency the employer used, so the medians mix currencies.

Remote jobs with pandas (needs pip install pandas pyarrow huggingface_hub):

import pandas as pd

jobs = pd.read_parquet(
    "hf://datasets/HeadcountLabs/job-postings/data/jobs/part-202610-s0-0.parquet",
    columns=["job_id", "title", "employer_name", "work_arrangement", "posted_at", "posting_url"],
)
remote = jobs[jobs["work_arrangement"] == "remote"]
print(f"{len(jobs):,} jobs in this file, {len(remote):,} remote")
# 315,523 jobs in this file, 6,995 remote

Streaming with Hugging Face datasets (no full download):

from datasets import load_dataset

jobs = load_dataset("HeadcountLabs/job-postings", "jobs", split="train", streaming=True)
for row in jobs.take(3):
    print(row["title"], "|", row["employer_name"], "|", row["posted_at"])

Live API

The dataset is a snapshot. The API serves the same jobs live, with job descriptions and search by title, place, employer, pay and date. Create a free key at api.headcountlabs.com/data/signup, then:

curl "https://api.headcountlabs.com/v1/jobs?q=registered+nurse&state=NY&limit=25" \
  --header "Authorization: Bearer $HEADCOUNTLABS_API_KEY"

curl "https://api.headcountlabs.com/v1/jobs/count" \
  --header "Authorization: Bearer $HEADCOUNTLABS_API_KEY"

Every organization gets 100,000 records free each calendar month, with no card. After that, records cost a flat $0.01 per 1,000. Counts are free. Details: headcountlabs.com/pricing.

MCP server

Claude, ChatGPT, Codex, Cursor and other MCP clients can search, count and read jobs through the hosted MCP server. Try it with no account:

claude mcp add --transport http headcountlabs-public https://api.headcountlabs.com/mcp/public

Or count every open job with one request:

curl -s https://api.headcountlabs.com/mcp/public \
  -H "content-type: application/json" -H "accept: application/json, text/event-stream" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"count_jobs","arguments":{}}}'

The signed-in endpoint is https://api.headcountlabs.com/mcp. Setup for each client: headcountlabs.com/mcp and the headcountlabs-mcp repository. For Claude Code there is also a plugin.

Job lists on GitHub

Open job lists built every morning from the API. Each job was still listed on the employer's own board within the past three days, and closed jobs drop off at the next update.

How to cite

Credit "Headcount Labs (https://headcountlabs.com)" when you use the data. GitHub's "Cite this repository" button reads CITATION.cff. BibTeX:

@misc{headcountlabs_job_postings_2026,
  author       = {{Headcount Labs}},
  title        = {Headcount Labs Job Postings},
  year         = {2026},
  month        = oct,
  publisher    = {Hugging Face},
  howpublished = {\url{https://huggingface.co/datasets/HeadcountLabs/job-postings}},
  note         = {Snapshot of 2026-10-19, ODbL 1.0}
}

License

The dataset and the sample in this repository are under the Open Database License (ODbL) 1.0. Our own values in them are under the Database Contents License (DbCL) 1.0. The example code in examples/ is under the MIT License. Job description text is the employer's own text and is not licensed by us. Employer names and logos belong to the employers that published them. Location data includes data from OpenStreetMap contributors (ODbL) and Overture Maps.

Questions, corrections or takedown requests: headcountlabs.com/contact or [email protected].

from github.com/HeadcountLabs/job-postings

Установка Job Postings

У этого сервера нет опубликованного пакета — он собирается из исходников. Открой репозиторий и следуй инструкции в README.

▸ github.com/HeadcountLabs/job-postings

FAQ

Job Postings MCP бесплатный?

Да, Job Postings MCP бесплатный — установка в пару кликов через Unyly без оплаты.

Нужен ли API-ключ для Job Postings?

Нет, Job Postings работает без API-ключей и переменных окружения.

Job Postings — hosted или self-hosted?

Self-hosted: сервер запускается локально на твоей машине командой из раздела установки.

Как установить Job Postings в Claude Desktop, Claude Code или Cursor?

Открой Job Postings на unyly.org, выбери вкладку своего клиента (Claude Desktop, Claude Code, Cursor) и нажми Install — конфиг сгенерируется автоматически, без правки JSON.

Похожие MCP

Compare Job Postings with

Не уверен что выбрать?

Найди свой стек за 60 секунд

Автор?

Embed-бейдж для README

Похожее

Все в категории data