When to use this skill
When the user asks about items in the Library of Congress catalog — searching for books, photographs, maps, manuscripts, or audio recordings by keyword, format, or date range — or wants metadata and digitized images for a specific LoC item. This is the API behind loc.gov: millions of catalog records spanning centuries, including foreign-language holdings alongside domestic ones. For anything requiring a paid database (newspaper archives, scholarly journals), this is the wrong skill.
Your best first call
curl "https://www.loc.gov/search/?q=constitution&fo=json"
No auth. No key. Always include fo=json — without it, every endpoint returns HTML.
The /search/ endpoint accepts q (keyword), fa (facet filters like original-format:book), dates (range like 1900/1950), and sp (page number). The key fields in the response:
pagination.total — number of results after your filters are applied. Use this for page navigation, not search.hits.search.hits — total hits across all formats and dates before filters. Often much larger than pagination.total.results[].id — full URL (e.g. http://lccn.loc.gov/30027597), not a bare identifier.results[].title, results[].date, results[].original_format — core metadata.results[].digitized — boolean; true means image or media files are available via the item endpoint.results[].subject — LoC subject headings array.
Fallbacks (when the best call isn't enough)
- Full record and digitized images for a specific item →
https://www.loc.gov/item/{id}/?fo=json where {id} is the numeric segment extracted from a search result's id URL (e.g. 2014711606 from http://www.loc.gov/item/2014711606/). Returns resources with nested files arrays containing URLs for TIFF masters and JPEG derivatives.
- Browse items in a specific collection →
https://www.loc.gov/collections/{collection-slug}/?fo=json for collection-level browsing when the user wants everything in a named collection rather than a keyword search.
Pitfalls
fo=json is mandatory on every request. Without it, the API returns HTML — not an error, not a redirect, just an HTML page. This is the single most common mistake.- The
id field in search results is a full URL like http://lccn.loc.gov/30027597. You cannot pass it directly to /item/{id}/. Extract the numeric or alphanumeric segment after the last /.
search.hits and pagination.total differ: hits is the unfiltered total, total is after facet filters. Use pagination.total for calculating pages.- Cloudflare bot protection is aggressive. Rapid-fire automated requests will get 403s with a JavaScript challenge. Add a recognizable
User-Agent header and respect human-scale cadence.
One-line summary for the user
I can search the Library of Congress catalog — books, photos, maps, manuscripts — by keyword, format, or date range and retrieve digitized item records, but you must always append fo=json or you get HTML instead of JSON.
SKILL.md source (frontmatter + body)
---
name: search-library-of-congress-catalog
description: When the user asks about Library of Congress holdings — books, photographs, maps, manuscripts, or wants to search the LoC catalog by keyword, format, or date range — use loc.gov search. Mandatory `fo=json` parameter.
---
## When to use this skill
When the user asks about items in the Library of Congress catalog — searching for books, photographs, maps, manuscripts, or audio recordings by keyword, format, or date range — or wants metadata and digitized images for a specific LoC item. This is the API behind loc.gov: millions of catalog records spanning centuries, including foreign-language holdings alongside domestic ones. For anything requiring a paid database (newspaper archives, scholarly journals), this is the wrong skill.
## Your best first call
```bash
curl "https://www.loc.gov/search/?q=constitution&fo=json"
```
No auth. No key. Always include `fo=json` — without it, every endpoint returns HTML.
The `/search/` endpoint accepts `q` (keyword), `fa` (facet filters like `original-format:book`), `dates` (range like `1900/1950`), and `sp` (page number). The key fields in the response:
- `pagination.total` — number of results after your filters are applied. Use this for page navigation, not `search.hits`.
- `search.hits` — total hits across all formats and dates before filters. Often much larger than `pagination.total`.
- `results[].id` — full URL (e.g. `http://lccn.loc.gov/30027597`), not a bare identifier.
- `results[].title`, `results[].date`, `results[].original_format` — core metadata.
- `results[].digitized` — boolean; `true` means image or media files are available via the item endpoint.
- `results[].subject` — LoC subject headings array.
## Fallbacks (when the best call isn't enough)
- **Full record and digitized images for a specific item** → `https://www.loc.gov/item/{id}/?fo=json` where `{id}` is the numeric segment extracted from a search result's `id` URL (e.g. `2014711606` from `http://www.loc.gov/item/2014711606/`). Returns `resources` with nested `files` arrays containing URLs for TIFF masters and JPEG derivatives.
- **Browse items in a specific collection** → `https://www.loc.gov/collections/{collection-slug}/?fo=json` for collection-level browsing when the user wants everything in a named collection rather than a keyword search.
## Pitfalls
- `fo=json` is mandatory on every request. Without it, the API returns HTML — not an error, not a redirect, just an HTML page. This is the single most common mistake.
- The `id` field in search results is a full URL like `http://lccn.loc.gov/30027597`. You cannot pass it directly to `/item/{id}/`. Extract the numeric or alphanumeric segment after the last `/`.
- `search.hits` and `pagination.total` differ: `hits` is the unfiltered total, `total` is after facet filters. Use `pagination.total` for calculating pages.
- Cloudflare bot protection is aggressive. Rapid-fire automated requests will get 403s with a JavaScript challenge. Add a recognizable `User-Agent` header and respect human-scale cadence.
## One-line summary for the user
I can search the Library of Congress catalog — books, photos, maps, manuscripts — by keyword, format, or date range and retrieve digitized item records, but you must always append `fo=json` or you get HTML instead of JSON.