Datasets

Overview

A dataset is a logical collection of labor market data that Lightcast MCP tools use to answer analytical questions.

Before running an analysis, you select a dataset that best matches the type of information you want to explore. The dataset determines which tools, dimensions, metrics, and filters are available for your request.

Datasets are product-level abstractions. They are not database tables and do not directly correspond to a single API or backend data source. Instead, each dataset provides a curated analytical experience with its own supported tools, metrics, dimensions, and response formats.

Dataset support is tool-specific. Before building an analysis request, use tool_info to view the supported dimensions, metrics, filters, and schema requirements for the selected dataset and tool combination.

Available Datasets

The following datasets are currently available in Lightcast MCP:

Dataset	Description	Supported Tools
`job_postings`	Real-time labor demand from deduplicated online job advertisements. Includes hiring intent, required skills, qualifications, employer and industry context, geography, time, and compensation signals.	`count`, `distribution`, `estimate`, `list`, `timeseries`, `top`
`profiles`	Labor supply and workforce qualification signals from professional profile records. Includes work history, skills, education, employer affiliation, geography, and estimated wage signals.	`count`, `distribution`, `estimate`, `top`
`core_lmi`	Government-sourced labor market data covering occupational employment, unemployment, labor force conditions, industry activity, earnings, and workforce measures. Includes `government_occupations`, `government_unemployment`, and `government_industry.`	`distribution`, `estimate`, `timeseries`, `top`

Why Datasets Exist

Datasets provide a consistent analytical experience across different labor market domains.

Benefits	Description
Tool routing	The dataset tells Lightcast MCP which tool behavior and data domain to use.
Schema validation	Ensures that only supported dimensions, metrics, and filters are used.
Concept normalization	Provides standardized dimensions, metrics, and taxonomies for each data domain.
Error prevention	Prevents unsupported dataset and tool combinations from being executed.

Datasets vs. Physical Sources

A physical source is where data originates. A Lightcast MCP dataset is the customer-facing analytical surface exposed through MCP tools.

Physical-source Thinking	Dataset Thinking
"Which backend API should I call?"	"What question am I trying to answer?"
Request formats vary by source.	Request formats are standardized through MCP tools.
Taxonomies and metrics may be source-specific.	Taxonomies and metrics are exposed as normalized MCP concepts.

Choosing a Dataset

Use the following guidance when selecting a dataset:

Question Type	Recommended Dataset
Hiring demand, employer posting activity, advertised skills, posting salaries, or demand trends	`job_postings`
Workforce supply, professional skills, education, employer affiliation, or estimated profile wages	`profiles`
Occupational employment, earnings, growth, openings, hires, or separations	`government_occupations`
Employment, unemployment, labor force size, or labor force participation	`government_unemployment`
Industry employment, compensation, establishments, hires, separations, or job creation and destruction	`government_industry`

Dataset-Specific Requirements

Some datasets support only a subset of tools, dimensions, or analytical workflows.

Dataset	Requirement or Limitation
`job_postings`	Analysis tools require a `time` filter.
`profiles`	Supports a `time` filter based on profile last-updated dates, but no required filter dimension is documented for all tools.
`government_occupations`	`timeseries` requires a `time` filter and yearly intervals.
`government_unemployment`	Industry and occupation filters are limited to supported classification levels.
`government_industry`	Supports `estimate`, `timeseries`, and `top` only.

Next Steps

After selecting a dataset, review the available dimensions, metrics, and tool schemas before building your request.

See Dimensions, Metrics, and Tool Contracts for additional details.