US Copyright Office¶

Read-only access to the Copyright Public Records System — the US Copyright Office's search index over registrations (post‑1978 + digitized card catalog) and recorded documents (transfers, assignments, licenses). The connector wraps the undocumented JSON API behind the web UI; no authentication required.

Auth¶

None. The Copyright Office publishes the records search index for free. The server does reject HTTP/1.1 requests with a 500 — the client uses HTTP/2 automatically (BaseAsyncClient's HTTP2 = True).

Quick Start¶

from patent_client_agents.copyright import CopyrightClient

async with CopyrightClient() as client:
    response = await client.search("Mickey Mouse")
    for record in response.records:
        print(record.copyright_number_for_display, record.title_of_work)

    # Lookup by registration or by name
    by_title = await client.search_by_title("The Great Gatsby")
    by_name = await client.search_by_name("F. Scott Fitzgerald Estate")

Functions¶

Method	Description
`search(query, field, page, page_size, sort_order)`	Search across all fields (`field="keyword"`), titles, or claimant names
`search_by_title(title, page, page_size)`	Convenience wrapper for `field="title"`
`search_by_name(name, page, page_size)`	Convenience wrapper for `field="name"` (claimant)
`get_record(public_records_id)`	Lookup a single record by its public records ID

Response Shape¶

SearchResponse carries:

metadata — took_ms, hit_count, query, etc.
histogram — facet counts for type_of_record, type_of_work, registration_class, registration_status, system_of_origin, recordation_item_type
records — list of CopyrightRecord with title(s), registration number(s), claimants, dates, work type, etc.

Records originating in voyager are post‑1978; card_catalog records are the digitized pre‑1978 backfile. Some fields (e.g. SR‑prefix sound recordings) arrive as scalars instead of arrays — the model coerces those to one‑element lists rather than rejecting the row.

Examples¶

# Sort the most-recent registrations for a claimant
recent = await client.search("Disney", field="name", sort_order="desc", page_size=25)

# Pull a specific record
detail = await client.get_record("voyager_67890")
print(detail.registration_status, detail.first_published_date)

Stability Notes¶

The API is undocumented (no service contract). Response shape has been stable through the migration to publicrecords.copyright.gov (the legacy cocatalog.loc.gov web UI is the same backend).
Detail pages return through the search index — there is no dedicated detail endpoint, so get_record searches by ID and filters.
HTTP/2 is non-negotiable: the server returns 500 on HTTP/1.1.

MCP Tool Surface¶

Two tools register unconditionally (no env gate — Copyright Office is public): search_copyright, get_copyright_record.