Output

The record envelope, JSON and Markdown, and the data tree on disk.

ant speaks JSON by default. Every record comes back in the same envelope, so one shape drives get, links, export, and graph no matter which site the record came from.

The envelope

{
  "@id": "goodreads://book/2767052",
  "@type": "goodreads/book",
  "@fetched": "2026-06-14T09:31:00Z",
  "@links": { "author": ["goodreads://author/153394"] },
  "data": { "title": "The Hunger Games", "...": "..." }
}

@id is the canonical URI of the record.
@type is scheme/authority, the site and record type.
@fetched is when ant dereferenced it.
@links groups the outbound edges by the field they came from. These are the same URIs links prints and export --follow walks.
data is the site's record, untouched.

It pipes straight into jq:

ant get goodreads://book/2767052 | jq '.["@links"]'
ant ls x://user/nasa -n 5 | jq '.[].["@id"]'

Markdown bodies

Records that carry prose tag it as the body. cat prints just that, and -o md renders it as Markdown wherever a body exists, falling back to JSON otherwise:

ant cat x://status/20            # the post text
ant get goodreads://book/123 -o md   # the description as Markdown

export --md writes a companion .md file with JSON front-matter next to each record that has a body.

The data tree

export materializes records under --data (default $HOME/data) so that a record's file path is its URI:

$HOME/data/goodreads/book/2767052.json
$HOME/data/goodreads/author/153394.json
$HOME/data/x/user/nasa.json

ll lists what is already there by reading the tree back into URIs, and import reads one file back into the envelope. The tree is plain files: browse it, grep it, or check it into git.