Financial reports, as clean Markdown

Structured, searchable financial reports for listed companies across the US, China, the UK, Japan, Korea and Taiwan — delivered through a simple REST API.

Want the full Markdown text, chapter-by-chapter, without limits? Get an API key

24,400+
listed companies
16
exchanges
910K+
Financial report documents
Markdown
Clean, structured format

Built for developers who need the full reports

Official disclosure sources

Every document is sourced from official regulatory disclosure platforms in the US, China, the UK, Japan, Korea and Taiwan.

Clean, structured Markdown

Raw PDFs are parsed into readable Markdown — headings, paragraphs and tables preserved, ready for analysis.

Precise chapters, not whole reports

Pull just the MD&A or risk section instead of the whole filing, saving tokens — with every figure traceable back to its source block.

Symbol-aligned with FMP

Query by ticker just like you already do — AAPL, 600519.SS, 7203.T, 005930.KS, 2330.TW. Drops straight into existing workflows.

One simple REST API

Authenticate with a single header, pull a list, fetch full text, or batch-download by document ID.

MCP server for AI agents

Connect Claude, Cursor, OpenAI Codex or DeepSeek via Model Context Protocol — ask in plain language, no code.

Try it in one line

$ curl "https://api.datasink.ing/documents?symbol=600519.SS&order=desc&size=1&\
with_content=1&apikey=YOUR_KEY"

Returns Kweichow Moutai's latest report as full-text Markdown.

Frequently asked

How do I get a Chinese A-share annual report as Markdown?
Send a GET request to the documents endpoint with the symbol and with_content=1, e.g. curl "https://api.datasink.ing/documents?symbol=600519.SS&with_content=1&apikey=YOUR_KEY". It returns Kweichow Moutai's filings as full-text Markdown, each one prefixed with a YAML frontmatter block (stock code, company name, report period, document type, source). A free API key is available by email at POST https://api.datasink.ing/free-key — no signup required.
Which markets and stock exchanges are covered?
Six markets and thirteen exchanges: the US (NYSE, NASDAQ, CBOE), China (Shanghai SSE, Shenzhen SZSE, Beijing BSE), the UK (London Stock Exchange), Japan (Tokyo TSE), South Korea (KOSPI, KOSDAQ, KONEX) and Taiwan (TWSE, TPEx) — about 24,400+ listed companies and 910K+ documents. All documents come from the official disclosure platforms: SEC EDGAR for the US, cninfo.com.cn for mainland China, filings.xbrl.org for the UK, EDINET for Japan, DART for Korea and MOPS for Taiwan.
Are these full-text reports, or extracted financial data?
Full text. Every filing is parsed from the official disclosure PDF into Markdown with headings, paragraphs and tables preserved, and unit notes kept inline (for example “> 单位:元”) so an LLM cannot mistake millions for units. The narrative sections — management discussion and analysis, risk factors, related-party transactions, notes to the financial statements — are the point; structured-number APIs are a different product and a useful complement.
Can I retrieve just one section, such as the MD&A, instead of a 300-page report?
Yes. GET /documents/{id}/sections lists every heading in a report, and GET /documents/{id}?section=%E7%AE%A1%E7%90%86%E5%B1%82%E8%AE%A8%E8%AE%BA%E4%B8%8E%E5%88%86%E6%9E%90 (that %-string is the heading 管理层讨论与分析, URL-encoded — non-ASCII headings must be encoded) returns just that section. This keeps RAG chunks small and avoids spending a context window on the whole document.
Which report types are included, and how far back does the history go?
Annual reports, semi-annual reports, Q1 and Q3 quarterly reports, and amendments or supplements. Coverage reaches back to roughly 1991 for the US, 2000 for mainland China, 2005 for South Korea, 2006 for Taiwan, 2015 for Japan, and 2020 for the UK. Every company page on this site lists exactly which periods are available for that ticker.
Can an LLM or agent call DataSinking directly?
Yes. An MCP server is published as the datasinking package (pip install "datasinking[mcp]", six tools including chapter-level retrieval), with a remote endpoint at https://api.datasink.ing/mcp. An OpenAPI specification is served at https://api.datasink.ing/openapi.json, so any tool-calling model can be pointed at the API without writing a client.
How can I report a problem or give feedback?
Use the chat assistant in the bottom-right corner of any page, we will read it.