DiligenceSquared

Fetch the full document behind an id. Accepts ids from search (transcript:<uuid>) and the fetch_id fields the report getters return (exhibit:<uuid>), plus qa:<uuid> and report:<project uuid>. Transcripts return windowed content; the report returns its curated payload. An exhibit returns its description plus, for chart/visual exhibits, the visualization source — a React component with the numbers, labels, and any quotes inlined (the same content the report shows). For verbatim expert quotes behind an exhibit, drill into the Expert Q&A (get_expert_qa_answers) or transcripts (search / fetch transcript:<id>).

fetch

Read every expert's answer to one Expert Q&A question. Returns per-transcript rows with the expert reference (byline), short and detailed answers, and the supporting transcript quote. Page with limit/offset for questions answered across many calls.

get_expert_qa_answers

Read the project report (client view): sections, the questions under each, and exhibit references (headline + fetch_id) — grouped by parent question. Pass question_ids (from get_report_outline) to fetch specific questions; omit for the whole report. Exhibits come as stubs with a fetch_id — call fetch("exhibit:<id>") for an exhibit's description + visualization source (React component with the data inlined). Large reports set `truncated` + `remaining_question_ids`; page with question_ids. Omit published_version_id for the latest published version (what the client sees); pass one from list_report_versions to read a specific version.

get_report

Browse the report's structure before pulling content. Returns sections (scopes) → parent questions → the exhibits under each (headline + fetch_id), with no narrative text. Use it to navigate, then get_report(question_ids=[...]) for the full content, or fetch("exhibit:<id>") for an exhibit's description + visualization source. Omit published_version_id for the latest published version (what the client sees); pass one from list_report_versions to browse a specific version.

get_report_outline

Read a transcript's content (windowed). Returns up to max_chars characters starting at offset, with `content_total_chars` and `next_offset` for paging through long transcripts. Content reflects your project's anonymization settings.

get_transcript

List the Expert Q&A questions available for a project. Each entry has a subquestion_id — pass it to get_expert_qa_answers to read every expert's answer to that question.

list_expert_qa_questions

List the DiligenceSquared projects your account can access. Returns each project's id (used by every other tool), name, industry, and which surfaces are available: transcripts are always available; expert_qa_available / report_available reflect whether those deliverables have been published to clients yet.

list_projects

List the published report versions the client can see (multi-version). Each is a label (never an internal version number) with its id and whether it is the default — the version get_report / get_report_outline read when no published_version_id is given. Pass an id back to those tools to read a specific version. An empty list means nothing is published to the client yet.

list_report_versions

List expert call transcripts for a project. Returns transcript metadata (id, reference/byline, name, call date) — use get_transcript to read content. Page with limit/offset; `total` is the full count.

list_transcripts

Semantic search over your projects' expert-call transcripts. Finds the passages most relevant to `query` across many transcripts and returns them ranked, each with a fetch id. This is the tool for "what did experts say about X" — open-ended, thematic, cross-call questions. For the other surfaces, use the structured getters (the progressive pattern): Expert Q&A via list_expert_qa_questions → get_expert_qa_answers; the report via get_report_outline → get_report; a full transcript via get_transcript. Exact competitor/vendor names are often better reached via the Expert Q&A getters than semantic search. Breadth-first: `max_transcripts` (default 50) controls how many transcripts are covered; `max_chunks_per_transcript` (default 3) how much of each. Scope to one project with project_id, or omit to search across your accessible projects (up to 10; projects_truncated flags when more exist). Page with offset/next_offset — the way to read past the host's result-size cap.

search