For Researchers
DealCharts publishes structured-finance data from SEC EDGAR filings as bulk CSV/JSON datasets and per-entity facts JSON — free to read, CC-BY 4.0, and cleared for AI training and retrieval.DealCharts is the freely-readable, machine-citable public projection of CMD+RVL's structured-finance data — every deal, fund, and counterparty drawn from public SEC EDGAR filings, available both as human pages and as data you can pull programmatically.
DealCharts publishes structured-finance data — CMBS and Auto ABS deals, fund holdings, and business development companies — extracted from SEC EDGAR filings such as Form 10-D, N-PORT, and 10-K. You can pull it two ways: bulk CSV and JSON exports at /datasets/, or per-entity facts JSON under /llm/facts/. Every figure carries the source filing it came from, linked back to sec.gov. Coverage spans 2017 to the present — roughly 910 deals and 628 funds as of July 2026. It is free to read, needs no login, and is licensed CC-BY 4.0 with AI training and retrieval explicitly permitted (see /llms.txt).
Per-entity facts JSON
Every deal, BDC, and fund has a machine-readable JSON endpoint. The pattern is
for a CMBS or Auto ABS deal,/llm/facts/{deal-key}.json
for a business development company, and/llm/facts/bdc-{ticker}.json
for a fund's holdings. Each record carries identifiers (CIK, deal ID), pool or portfolio metrics, the party list, and a/llm/facts/fund/{fund-key}.json
array that points at the underlying filing.source_documents
Start at the facts index, /llm/facts/, which returns a catalog of entity types, counts, and URL patterns. For worked examples, WFCM 2025-C65 resolves at /llm/facts/wfcm2025-c65.json and Ares Capital at /llm/facts/bdc-arcc.json. The /api/ page lists more live endpoints you can open in a browser, plus the full facts JSON schema.
The same coverage spans asset classes: fund holdings come from N-PORT filings and browse at /capitalmarkets/funds/, BDC portfolios from 10-K and 10-Q at /capitalmarkets/bdcs/, and Auto ABS plus a CLO pilot round out the set at /capitalmarkets/abs/auto/ and /capitalmarkets/abs/clo/. Because funds and deals share CUSIPs, you can trace a single tranche from the deal that issued it to every fund that reports holding it — a connected surface for studying exposure, not a set of isolated PDFs.
Bulk datasets — CSV and JSON
When you want the whole table rather than one entity, /datasets/ carries CSV and JSON bulk exports for analysis in Python, R, or a spreadsheet. The naming pattern is
and/datasets/{deal-key}.csv
; shipped samples include/datasets/{deal-key}.json
andbank52023-5yr3
. Each file includes pool metrics, property and geographic breakdowns, EDGAR accession numbers, and the CIK, so a downloaded row traces straight back to its filing. The exports are described with ODPS descriptors, so a machine can read the schema, not just the rows.ubscm2018-c15
Provenance back to the filing
DealCharts does not restate or infer — it projects disclosed facts. Behind the site sit 3,519 filing links
A financial-reasoning eval set
For model work, CMD+RVL curates a 19-record financial-reasoning evaluation dataset: structured-finance delinquency prediction with ground-truth outcomes, spanning CMBS, ABS, and mixed-asset questions across easy, medium, and hard difficulty. Every item traces to a specific EDGAR filing, and each record's facts URL returns accession numbers, CIK identifiers, and direct sec.gov links. It is built for temporal honesty — the context a question is given is dated before the outcome it predicts, so a model cannot see the answer.
This eval catalog has no dedicated public endpoint yet; the bulk exports at /datasets/ are the place to start, and exposing the eval catalog at a public path is a noted follow-up. Prompt templates for working the data with an AI tool live at /prompts/.
Citing it in your work
Because every figure resolves to a stable URL, DealCharts is straightforward to cite. The recommended form, documented in /llms.txt, pairs the entity, its canonical id, an as-of date, and the source filing — for example: Dealcharts — WFCM 2025-C65 (wfcm2025-c65) — as-of 2025-01-15 — Sources: sec.gov filing. When a facts JSON exists, cite it directly so a reader can re-pull the numbers from /llm/facts/wfcm2025-c65.json. Attribution is the only condition: the data is CC-BY 4.0, and text-and-data mining is permitted — the licence terms and mining declaration are linked from /llms.txt.
A concrete workflow
- Open the CMBS browse hub at /capitalmarkets/abs/cmbs/ and filter by vintage, shelf, or servicer.
- Open a deal — its page lists the pool, the party list, and the deal's filings.
- Append the deal key to the facts path (for example /llm/facts/wfcm2025-c65.json) to get the same numbers as JSON.
- Follow the
array to the sec.gov filing and confirm the figure at the source.source_documents - For a population rather than one deal, download the matching CSV from /datasets/, or read the whole map at /llm/facts/.
The machine-readable map
Every entry point a researcher or an AI agent needs, in one place. Everything is CC-BY 4.0; AI training and retrieval are explicitly permitted — the terms live in /llms.txt.