The labour market, as records not opinions
Indices
Ghost Jobs Index - monthly. How long the postings on employers' own career sites have been up, and how many carry ghost-job signals, by role family, country, employer size, seniority, work mode and pay transparency. Measured from the boards, not surveyed; every edition is frozen at its own address and published under CC BY 4.0 with CSV and JSON downloads.
Global South Remote Hiring Index - monthly. Of the fully remote jobs on employers' own career sites, which ones a person living in Africa, Latin America or developing Asia may actually apply to, by country, region, role family and seniority, and who the employers are. Read from the eligibility each posting states; frozen editions under CC BY 4.0, every table on Hugging Face.
The datasets
| Dataset | What a row is | Updated |
|---|---|---|
| Job postings | One opening read from an employer's ATS board or career page: employer, role and role family, seniority, employment type, work mode and remote scope, locations with ISO country codes, stated or estimated pay, required technologies, visa and equity flags, first seen, last confirmed, lifecycle journal, trust verdict, source URL. | Continuously; boards refetched on a per-employer cadence |
| Company hiring signals | One employer: identity and verified domain, industry, size band, founded year, headquarters and offices, ATS vendor, technology stack seen in postings, funding rounds, open roles, roles added and closed over 30 and 90 days, hiring velocity and a weekly history. | Daily snapshots |
| Salary benchmarks | One cell of role x seniority x geography: percentiles of stated pay, the calibrated estimate, the sample size behind it and the method that produced it. | Rebuilt as the posting corpus moves |
| AI agents catalogue | One agent, MCP server, framework or skill indexed from GitHub, the MCP registry and Hugging Face Spaces, described in plain language: what it does, who it is for, what it needs, skill level, pricing. | Continuously |
How each field is produced - the re-check window, the stale and ghost thresholds, how an estimate is computed and when it stays empty - is written out in the methodology.
Samples and dictionary
Free files, no account, no form:
- Job postings sample (CSV)
- Company hiring signals sample (CSV)
- Salary benchmarks sample (CSV)
- Data catalogue overview (PDF)
- Data dictionary (PDF) - every field, its type and its meaning
The same samples, with the dictionary, are on Hugging Face: alion-job-market-samples; the Ghost Jobs Index editions: ghost-jobs-index.
Datasets on Hugging Face
Monthly samples for research and models, in the alion-io organization; every row links back to its page here:
- ghost-job-detection - 20,000 postings with real lifecycle labels: still up five weeks later, or closed
- job-posting-extraction - 8,000 posting texts with the fields Alion extracts (role, seniority, remote scope, salary, technologies)
- salary-benchmarks - pay percentiles by role family, seniority and country
- whos-hiring - the 1,000 employers opening the most roles this month
- role-technology-taxonomy - the role and technology trees, roles in eleven languages
- remote-hiring-global-south - the Global South Remote Hiring Index tables: countries, regions, employers, role families, one file per month
Query it directly
The same records are served live to software. POST https://alion.io/mcp is a stateless MCP server (Streamable HTTP, JSON-RPC 2.0); its read-only tools for these records - search_jobs, get_job, resolve_company, search_companies, get_company - also exist as plain HTTP endpoints under /mcp/rest/, described in OpenAPI 3.1 at /mcp/openapi. No account is needed to start: https://alion.io/mcp/open answers search_jobs, get_job, search_companies, similar_companies and resolve_company free (compact lists, 300 calls a day per network), and any other call can be paid per call in USDC over x402. With an Alion account (an API key or a sign-in) calls are paid from its credits. Setup for Claude, Cursor and other clients is on the developers page.
Coverage
Worldwide, with particular depth in Europe, the Gulf, Central Asia, India and remote-first hiring. Roles span software engineering, data and AI, infrastructure, security, hardware and chip design, robotics, product, design, analytics and technical leadership. Postings written in other languages are normalised into one English vocabulary for roles, seniority, employment type and work mode, while the original text is kept.
Using and citing it
The sample files and the API responses may be used, including commercially, under the terms of service, provided the source is attributed. A citation that lets a reader check the number looks like this:
Alion technology labour-market data, alion.io, retrieved 11 November 2026.
If you are quoting a figure that moves - live roles, employers hiring, a salary percentile - say when you took it, because it will have moved by the time anyone re-reads the sentence. For a bespoke extract, a historical slice or a licence beyond the terms, write to us.