ElevenFlo MCP exposes 13 read-only bankruptcy research tools. The live
tools/list response is the exact input and output contract. Inputs reject
unknown fields; tool errors use {code, message, retryable?, recovery?}.
#Recommended research loop
find_casesto resolve acase_id.list_docket_entries,search_filings, orfind_case_document_hubsto select documents.search_document_chunksto find supporting text.read_document_chunksorextract_document_passagesfor exact language.search_document_summariesorget_document_summaryfor synthesis.
Public MCP uses case_id for cases and UUID document_id values for document
tools. Use identifiers exactly as returned; no compatibility aliases are
accepted. When a document row has has_primary_pdf: true, that UUID is also
the verified /api/search/open/<uuid>/ link handle.
#Case and taxonomy
#find_cases
Find cases by debtor identity or by structured case metadata. Case-type tags describe product workflow, not bankruptcy chapter.
- Inputs: optional
query,case_type,petition_date_from,petition_date_to, andassets_over_usd;limitis 1 to 50. At least one search criterion is required. - Returns: bounded case rows with
case_id, identifying metadata, petition date, disclosed petition asset range, coverage state, and case-type tags. Filter-only results are newest first.
#list_document_types
List canonical filing categories and tags.
- Inputs: optional
query. - Returns: category keys, labels, and canonical tags.
#list_docket_entries
List recent documents for one case. For confirmation, emergence, or current
status, first filter for confirmation_order and effective_date_notice; the
latest terminal notice controls over projected dates.
- Inputs:
case_id, optionallimitfrom 1 to 25 anddocument_type_tags. - Returns: bounded document rows, including
has_primary_pdfwhen known.
#Cross-document search
#search_filings
Search filing evidence across all cases or within one case.
- Inputs:
query; optionalcase_id,document_type_tags, andlimit. - Returns: bounded document rows with matched excerpts when available.
#search_news
Search bankruptcy news and source coverage.
- Inputs:
query; optionalcase_idandlimit. - Returns: bounded source rows with publisher metadata and snippets. Raw article text is not exposed.
#search_transcripts
Search hearing and court transcripts.
- Inputs:
query; optionalcase_idandlimit. - Returns: bounded transcript document rows.
All three search tools require every meaningful query term in matching evidence so a single common word cannot produce unrelated results.
#Summaries
#search_document_summaries
Search stored document summaries within one case.
- Inputs:
case_id,query, optionallimit. - Returns: bounded summary rows and their document IDs.
#get_document_summary
Read the stored summary for one document.
- Input:
document_id. - Returns: one summary or a typed not-found error.
#Exact text
#search_document_chunks
Find supporting chunks within selected documents.
- Inputs: one to 25
document_idsandquery. - Returns: matching chunk IDs, exact text, and offsets grouped by document.
#read_document_chunks
Read selected chunks from one document.
- Inputs:
document_idand one to eightchunk_ids. - Returns: exact chunk text and offsets.
#extract_document_passages
Extract passages answering a focused question from one document.
- Inputs:
document_id,query, optionalmax_chunksfrom 1 to 8. - Returns: selected exact chunks and offsets.
#Document relationships
#find_case_document_hubs
Find the most connected documents in a case citation graph.
- Inputs:
case_id, optionallimitfrom 1 to 25. - Returns: bounded document rows with citation counts.
#list_document_relationships
List citation relationships for selected documents.
- Inputs: one to 25
document_ids, optionaldirection, and optionallimit. - Returns: bounded source/related document pairs.
#Boundaries
The public MCP cannot write data, file documents, send messages, manage accounts, expose internal operating tools, or return raw publisher article text. See Permissions and data access.