About Clarity Atlas

Clarity Atlas is a public data platform configured for Atlanta Regional Commission. It holds 1,005 indicators for 1,582 geographies across 7 levels, every one with its source table, universe, vintage and margin of error. Staff add datasets, indicators, geographies, layers and dashboards in an administrative workspace without developer work.

The Regional Housing Dashboard and the Economic Mobility Dashboard are configurations of the same engine: curated indicators, goal records, county and city profiles and map layers that staff edit in the workspace.

Indicators. U.S. Census Bureau, American Community Survey 5-year estimates, data profile tables DP02 (social), DP03 (economic), DP04 (housing) and DP05 (demographic), vintage 2023. Values are loaded directly from the Census Data API with their published labels; counts and percentages are the Bureau’s own, never recomputed from smaller geographies. Margins of error are at the 90% confidence level.

Geographies. Census TIGERweb boundaries, generalized for the web: counties, incorporated places, ZIP Code Tabulation Areas, census tracts, block groups, congressional districts and state legislative districts within the eleven ARC counties (Cherokee, Clayton, Cobb, DeKalb, Douglas, Fayette, Forsyth, Fulton, Gwinnett, Henry, Rockdale).

Classification. Quantile, equal interval, natural breaks (Jenks) and custom breaks; the legend always states the method and class count. Dashboard regional figures sum counts across counties, recompute shares from counts, and use the median of county medians for medians.

Goals. Goal records on the dashboards carry a baseline, a target, a current value and a note. In this demonstration the targets are illustrative placeholders until the adopted strategy goals are entered.

Two capabilities, one rule: a model chooses what to show, the database decides what the numbers are. The “Ask the data” box turns a plain-language question into an indicator, a geography level and a place using a deterministic parser; a hosted model can replace the parser later without changing what it returns.

A Model Context Protocol server exposes the same functions to any MCP-capable assistant (Claude, ChatGPT, Microsoft Copilot and others): search the catalog, read metadata, fetch values for a level or an area, compare areas, and get a link that opens the exact map. Endpoint: /api/mcp (Streamable HTTP, no session required). Tools: about, search_indicators, describe_indicator, list_geographies, get_values, get_area, compare_areas, ask, map_link.

Every table and map exports with its metadata: CSV and Excel with an indicators sheet, GeoJSON with a metadata member, File Geodatabase through GDAL where the host has it installed, and PNG, JPG and PDF map images with the legend, method and source printed on them. Boundaries publish as GeoJSON at /api/geo/<level> and values at /api/values, so ArcGIS Online, QGIS and spreadsheets can read them directly.

Unlimited users, no viewer licenses. Accessibility statement.