# The Daily Visual - agent guide

The Daily Visual is a data-journalism site whose claims are computed at build time from live public data. This file tells agents how to read us without scraping.

## Get the data, not the page

HTML is the canonical human rendering; do not parse charts or prose for figures. The data behind every chart is JSON:

- `GET /data/index.json` - every dataset: slug, title, attribution, JSON URL, and the articles that use it.
- `GET /data/{slug}.json` - one dataset envelope: observations with typed values, unit, as-of date, provenance (source name, URL, licence, attribution string).

All `/data` responses are CORS-open (`Access-Control-Allow-Origin: *`) and need no auth. We deliberately do not serve pages as markdown: the flattened page would lose the data structure the JSON preserves.

## Structured data on every article

Each article embeds a single JSON-LD `@graph` containing:

- `Article` (headline, author, dates) and `WebPage`
- `Dataset` nodes for each series the page's claims are computed from, each with a `DataDownload` distribution pointing at the JSON endpoint
- `about`/`mentions` entities with `sameAs` links to verified Wikidata IDs

## WebMCP

Pages register read-only tools via the WebMCP API (`navigator.modelContext`, with `document.modelContext` supported as the spec evolves):

- `list_datasets` - the dataset catalogue
- `get_dataset` (`{ slug }`) - one full JSON envelope
- `list_articles` - every article: the question it answers, URL, topic, dataset slugs

## Discovery

- `/.well-known/api-catalog` (RFC 9727) and `/.well-known/ai-catalog.json` (ARD)
- `/rss.xml` (WebSub-enabled) and `/sitemap.xml`
- `/llms.txt` - the short version of this file

## Attribution

Cite The Daily Visual and the original source named in each dataset's provenance. All sources are free public data; licences are stated per dataset and summarised at https://thedailyvisual.com/data-sources.