MCP reference
get_document
Retrieve a document’s text, metadata and any available AI summary.
Arguments
| Parameter | Meaning |
|---|---|
| identifier · required | ECLI, ELI, CELEX or id from a search result |
| type | Optional: case_law or legislation; useful for ambiguous identifiers |
| language | Desired text language as an ISO code; defaults to German, then English, then a held language |
| cursor | Continuation token from text.next_cursor |
| max_characters | Requested chunk size, at least 1,000 characters; default 20,000, subject to the server limit |
| include_full_text | Request the largest possible text in one response; default false |
| anchor | Start at a text anchor, e.g. randnummer-42 |
| paragraph_from / paragraph_to | Select a range of official paragraph numbers |
Example: retrieve a known judgment
{
"identifier": "ECLI:DE:BVERFG:2019:RS20191106.1BVR001613",
"language": "de",
"max_characters": 20000
}Continue reading long texts
Check text.complete. If it is false, pass text.next_cursor back unchanged as cursor with the same identifier and selection. include_full_text does not bypass the server limit: further paging may still be necessary. The default maximum is 200,000 characters.
If a cursor is no longer valid, restart retrieval without cursor. status: fetching means that text is being loaded; wait retry_after_seconds and repeat the call.
Metadata and citations
available_languages lists held language versions. related_documents may list other documents in the same CJEU proceeding, such as the judgment and the Opinion. Paragraph numbers are retained as [Rn. 42]; text.paragraph_numbering describes the numbering. An ai_summary is machine-generated and separate from the original text.