Developer tools

Updated 16 October 2026 · free · open source

The Face Analysis publishes three free developer tools: an MCP server that lets AI assistants look up sourced face measurement statistics, the dataset behind those statistics as JSON and CSV, and an embeddable face shape widget for other websites. The code is MIT-licensed on GitHub, the packages are on npm, and the aggregate numbers are CC BY 4.0.

Packages

thefaceanalysis-mcp

MCP server for AI assistants: sourced face measurement statistics, face type rules, methodology and a face shape classifier. Seven read-only tools.

npx -y thefaceanalysis-mcp

npm GitHub Docs

thefaceanalysis-data

The face measurement statistics from 102 faces as JSON and CSV, with the type rules and every threshold. Data CC BY 4.0.

npm install thefaceanalysis-data

npm GitHub Docs

thefaceanalysis-embed

The free face shape finder as a web component and a React wrapper, for salons, opticians and beauty sites.

npm install thefaceanalysis-embed

npm GitHub Docs

What people use them for

AI assistants

Let Claude, ChatGPT, Codex or Cursor look up sourced face measurement statistics, read the exact face shape rules and classify a face shape from four distances, with every answer carrying its source.

MCP server and setup for each client

Researchers and analysts

Download the percentiles of fWHR, midface ratio, symmetry, canthal tilt and pupillary distance from 102 faces, with the type rules, every threshold and the dated change history. CC BY 4.0, JSON and CSV.

Face measurement statistics

Salons, opticians and beauty sites

Put the free face shape finder on your own page in two lines of HTML, or as a web component or React component. The visitor's photo is measured in their browser, with no sign-up.

Embed the widget

Where everything lives

What	Address
Developer docs	https://docs.thefaceanalysis.com
Docs for AI systems (llms.txt)	https://docs.thefaceanalysis.com/llms.txt
Hosted MCP server	`https://mcp.thefaceanalysis.com/mcp`
GitHub organization	https://github.com/The-Face-Analysis
npm: MCP server	https://www.npmjs.com/package/thefaceanalysis-mcp
npm: dataset	https://www.npmjs.com/package/thefaceanalysis-data
npm: widget	https://www.npmjs.com/package/thefaceanalysis-embed
Data files	https://thefaceanalysis.com/data/face-measurements.csv

What the data is and is not

Every number comes from 102 adult faces photographed in one city in one month. They describe those faces, not a population: no count is a prevalence and no median is a norm. The face type splits show how our detectors divide those faces, and we claim no accuracy figure for any type label. Our 1-10 score did not predict 2,513 people's attractiveness ratings of the same faces. Read the methodology before quoting a number, and see every method change in the changelog.