Canli Markets
io.github.arhancanliv0.2.1更新於 Oct 11, 2026
Markets, funds and the economy from the sources: options, shorts, filings, Fed, macro, prices.
概覽
讓助理以唯讀方式取得美國市場、基金、申報文件與宏觀經濟的公開資料,來源包括 SEC、財政部、FRED、世界銀行及行情來源。
- 功能
- 提供大量唯讀工具,涵蓋公開金融資料來源:公司概況與申報文件、申報全文檢索、財報行事曆、分部營收、內部人交易、基金與 13F 持股、大額持股人、放空資料、選擇權鏈、波動率曲面、期貨持倉、公債殖利率與標售、FRED 序列、FOMC 決議,以及世界銀行各國資料。另提供跨市場研究工具,例如公司篩選、一次呼叫產生公司報告、事件研究與申報用語趨勢。結果以表格形式回傳,並附上來源網址與文件證據,方便引用與核對。
- 適用情境
- 適合助理需要回答美國上市公司、基金、申報文件、選擇權、放空或宏觀資料的問題,而不想自行建置資料管線時。適用於需要可引用、可溯源數字的分析與研究情境,而非僅查詢單一公司。
- 執行需求
- 透過 stdio 在本機執行,通常使用 npx。預設無金鑰行情來源不需要帳戶。選用環境變數:SEC_USER_AGENT(姓名與電子郵件,用於 SEC 公平存取)、ALPACA_API_KEY_ID 與 ALPACA_API_SECRET_KEY,或 TIINGO_API_KEY 以改用其他美股行情來源。需要能連線至這些公開資料來源的網路。
安裝
在 SourceWeft 中
- 開啟 儀表板中的 Canli Markets,將其新增到工作區。
- 為需要使用其工具的對話啟用該服務。
Desktop only,透過 STDIO。 STDIO 服務會啟動本機處理程序,因此需要 SourceWeft 桌面主機。
其他 MCP 客戶端
參照 儲存庫 中的啟動說明。
README
canli-markets-mcp
Public market data for AI agents, read from the sources themselves: SEC EDGAR, the US Treasury, FRED, the World Bank and prices for stocks, currencies, futures, indices and crypto, with no key and no account. Every request goes from your machine straight to the source, and every result names the URL it came from, so an agent can cite it.
Set SEC_USER_AGENT to your name and email ("Jane Doe [email protected]"), as SEC's fair-access
policy asks. Prices need no key: they come from Yahoo Finance's public chart data, which is
unofficial and for personal use under Yahoo's terms. To use your own account instead, set
ALPACA_API_KEY_ID and ALPACA_API_SECRET_KEY (paper keys work) or TIINGO_API_KEY;
CANLI_KEYLESS_PRICES=0 turns the keyless source off.
Research across the whole market
Other finance servers answer one company at a time. These answer questions about the market:
Examples: "Which large companies grew revenue fastest in 2025, and what do they trade at?", "Rank companies over $50 billion of revenue by R&D intensity", "Does NVIDIA's stock rise after its earnings releases?", "How fast are filings adopting the phrase 'agentic AI'?".
Figures no real company reports (one 2026 filing tagged $1.157 billion of net income as $1,157 billion) are held out of rankings and listed apart, with the reason.
Every result also carries evidence: the documents read, with the SHA-256 of the exact bytes
received and when, so any figure can be checked against the same document later.
Tools
Examples an agent can now answer in one or two calls: "What are the new risk factors in Nvidia's latest 10-K?", "Did any Tesla insider buy shares on the open market this year?", "What did Berkshire Hathaway sell last quarter?", "Where is the 2s10s spread today?", "What is core PCE inflation year over year?".
Results are columns and rows, so a series feeds straight into analysis: with
canli-mcp, economic_series or price_history output
goes into any of its 235 quant tools in the same batch.
Checked against the source
The tests replay real SEC, Treasury and FRED responses and check values read off the documents: Berkshire Hathaway's 13F for the quarter ending 30 June 2026 totals $299,253,556,246 over 89 entries, exactly the figures on its cover page; Apple's 10-K risk factors come back as Item 1A alone; a Tesla Form 4 sale reads 2,605.75 shares at $360.134.
Limits
- Filings are shown as filed; nothing is edited or verified.
- Form 4 is due two business days after a trade; Form 5 filings are not read.
- A 13F lists long US-listed equity and option positions at quarter end, filed up to 45 days later; it omits shorts, cash and confidential positions. Amendments are not merged.
- FRED revises recent values of many series; some series are copyrighted by their sources.
- Keyless prices are Yahoo Finance's unofficial chart data, for personal use; they can change or stop without notice. Alpaca's free IEX feed reports IEX volume only.
- Not investment advice.
See SECURITY.md for exactly what is sent where.
來源:mcp-markets/README.md,提交 d2d0c99
工具
0版本歷史
1- v0.2.1最新Oct 11, 2026

