What you get with Institutional

Dragon-tiger list broker seats: the buy/sell seat breakdown (up to 10 broker branches) for one stock on one appearance, with a hot-money tag. data rows: name (branch) / total_buy (CNY) / total_sell (CNY) / net (buy − sell) / is_youzi (hot-money seat) / log_id. Top-level meta: code / name / date (appearance date) / qjzf (interval change %) / rows. ⭐ Who is buying — far more concrete than aggregate institutional stats: you see whether it is the Shanghai/Shenzhen Connect dedicated seat, a named brokerage branch, or a hot-money seat. ⚠️ The dragon-tiger list is post-close data: leave date empty for the most recent appearance, and trust the returned date. ⚠️ Up to ~10 seats per call (source limit). Paid tier — Institutional only.

Institutional participation board (whole market): every stock that institutional seats traded on the dragon-tiger list on a given session, in one call (~51 stocks per session). data rows (amounts in CNY): code / name / net_buy (institutional net buying — ⚠️ negative means institutions were net sellers) / join_num (seats involved) / change_pct (%) / turnover (CNY) / amplitude (%) / turnover_rate (%) / float_cap (CNY) / market_cap (CNY) / fengkou_plates (theme plate codes; plate names are not provided). Top-level meta (siblings of data): date (the actual session) / total / returned / sort / unit. ⭐ How to use it: sort=net_buy shows who institutions really bought today (negatives are sells); pair with /v1/lhb-seats to drill from a stock down to which broker branch bought it. ⭐ Versus /v1/lhb: that one is the split board (stocks listed by change / turnover criteria); this one keeps only the institution-involved names and gives the institutional amount directly — no manual cross-checking. ⚠️ The dragon-tiger list is post-close data: leave date empty for the latest session and trust the returned date. ⚠️ net_buy is a net figure (buy − sell) and can be negative (26 of the 51 rows were negative on the measured session) — ⛔ do not take the absolute value as buying volume. Paid tier — Institutional only.

Foreign (northbound) holdings by A-share sector, published quarterly; 20 sectors measured. data rows: plate_code / plate_name / raw (includes holding value columns). Top-level: date (the actual reporting period) / total_net (CNY, can be negative) / total_hold (CNY) / available_dates (7 quarter-ends). How to use it: total_net shows whether foreign money is net in or out; go sector by sector to see which line is being added. You must pass a quarter-end date (e.g. 2026-09-30); empty picks the latest quarter-end automatically, and a non-quarter-end date returns empty (not a fault). This is sector level, not per-stock holdings. Not the same as /v1/fund-south (Stock Connect southbound into HK) - this one is A-share northbound. Paid tier - Institutional only.

Which themes are printing 100-day new highs today, aggregated by theme. data rows: theme / plate_code / raw (includes the counts); top-level date. Themes where new highs cluster are where the money is working; compare with /v1/sector (change leaderboard). Paid tier - Institutional only.

How cold or hot the short-term tape is today: break rate plus a 9-trading-day series of money-making stats. Top-level date / break_count (limit-ups that broke) / break_rate (%). data rows (newest first): date / success_rate (%) / profit_rate (%) - can be negative. How to read it: success_rate sliding for several days = the tape is cooling; profit_rate turning positive = money-making is back. break_rate = broken / (broken + limit-ups) x 100; on 2026-10-23 that is 12 / (12 + 82) = 12.766%, verified digit for digit. Moves intraday (5-minute cache). Paid tier - Institutional only.

The answer to "why did this stock limit up": for each past limit-up, the full reason text plus the sector it belongs to and its "dragon rank of the day". data rows (newest first): date / type / dragon_rank (e.g. 日内龙一 = #1 of the day, empty = not a leader) / plates (sector codes) / reason (full text) / concept. ⭐ Versus /v1/limit-up: that endpoint lists today's whole market limit-ups with a one-line reason; this one gives one stock's history, so you can see which theme it limit-ups on every time. ⚠️ code must be an A-share (6 digits, optionally prefixed). Paid tier — Institutional only.

Which stocks had their limit-up smashed open on a given day. data rows: code / name / change_pct (%) / main_net (CNY) / amount (CNY) / turnover_rate (%) / concepts (e.g. fintech, financial concepts) / raw. Complements /v1/limit-up: that one shows the sealed boards, this one shows the broken ones - compare the two to judge the seal quality of that day. Empty date = the latest weekday (weekends fall back to Friday); pass an explicit date around public holidays. Paid tier - Institutional only.

Where late-session money is grabbing: about 292 stocks market-wide with grab amount, theme and change. data rows: code / name / fund_tag (e.g. fund, empty = none) / theme / change_pct (%) / chase_amount (CNY) / masked / raw. A few symbols are withheld and carry masked=true while the amount stays complete; do not treat those rows as tradable symbols. Results are re-sorted by sort (default chase_amount descending), so the first row is the largest grab of the day. Paid tier - Institutional only.

单股大单资金分时：每 1 分钟一个点，net = 该分钟大单净额（元），正数=大单净买入、负数=净卖出。

逐笔大单成交：单只票当日每一笔成交额 ≥ 阈值的大单，并按主动 / 被动 × 买 / 卖四方向标注。

Big-order pending orders: every order whose order volume or order amount exceeds the threshold, listed one by one, with limit-up and cancelled flags. data rows: time / order_id / price / volume (lots) / amount (CNY) / side (1 buy · 2 sell) / side_cn / is_limit_up (1 = placed at the limit-up price) / is_cancel (1 = cancelled). Top-level meta: vol_threshold (lots) / tur_threshold (10k CNY) / total / returned / truncated. ⭐ What it answers: is that limit-up seal real or staged — a big buy order that is posted then pulled (is_cancel=1) is a textbook fake seal. Pair it with /v1/bigorder-flow (what actually traded) to separate "posted but never filled" from "actually bought". ⚠️ Today only; about 200 rows per call max. ⚠️ This is the order (pending) side, not trades, and not a ten-level order book. Paid tier — Institutional only.

Per-day big-order net flow, up to 100 trading days back. data rows: date / net (that day's big-order net amount in CNY; positive = net buying, negative = net selling), newest first. Top-level fields (siblings of data): code / days / upstream_days / unit. ⭐ Use case: /v1/bigorder only covers today; this one looks back — is the big money accumulating steadily or just today's spike, and on which day did it turn. ⚠️ The source always returns 100 trading days; limit only truncates locally (saves bandwidth, not a call). ⚠️ net is the big-order metric (aggregated by trade amount) — not main-capital net flow, not the exchange dragon-tiger list, not tick data. ⚠️ Non-trading days return the 100 days ending at the last session — trust data[0].date. Paid tier — Institutional only.

Minute-level four-way main-capital split: main buying is split into active buy + passive buy, main selling into active sell + passive sell, one row per minute. Rows (amounts in CNY): time / trades / net / main_buy / main_sell (negative) / active_buy / passive_buy / active_sell / passive_sell. Top-level fields (siblings of data): day / points / mode (realtime | history) / summary (whole-day totals) / buy_avg_price / sell_avg_price (volume-weighted average price of main buying / selling). ⭐ How to read it: high active_buy = someone is sweeping the offer; high passive_buy = your bids are being hit; concentrated active_sell = active distribution; heavy passive_sell = offers being lifted. Only the four-way split tells you who is hitting whom. ⭐ Versus /v1/bigorder-flow: flow is the tick-by-tick trade list (which specific trade; today only, ≤200 per call); this endpoint is the minute aggregation (rhythm + average price, history supported, no row cap). ⚠️ Pass date for a past session; omit for today. ⚠️ This is a capital-flow metric (split by trade direction) — not an order queue, not a ten-level order book. Paid tier — Institutional only.

Whole-market main-capital leaderboard (~5,500 stocks in one call), 8 sort modes, pagination and historical lookup. sort: net_in (default) / net_out / buy / sell / gain / loss / turnover / mcap. Rows (amounts in CNY): code / name / price / change_pct / main_buy / main_sell (negative) / main_net / turnover_rate / amount / market_cap / concept. Top-level fields (siblings of data): total / returned / sort / date / offset / truncated / unit. ⭐ Self-check identity: main_buy + main_sell == main_net (holds for 100% of all ~5,500 rows). ⭐ Drill-down path: pick a stock here → /v1/bigorder (minute net flow) / /v1/bigorder-split (four-way split). ⚠️ limit up to 6000 (whole market in one call, ~1 MB); use offset to page. ⚠️ market_cap is the raw source market-cap field — for sorting / coarse screening only; do not mix it with market-cap fields from other endpoints. ⚠️ This is a capital-flow metric (aggregated by trade amount and direction) — not the exchange dragon-tiger list, not seat data, not ticks. Paid tier — Institutional only.

Lock an API key to a fixed set of IPs, so a leaked key cannot be used from anywhere else. GET returns the current list · POST replaces it (ips=1.2.3.4,5.6.7.8, up to 50) · DELETE clears it (the key is then usable from any IP again). When the list is non-empty, requests from other IPs get 403. Authenticate with the same key you are locking. ⚠️ Calls from your own current IP always pass, so you cannot lock yourself out; use DELETE to recover. Paid tier — Institutional (¥199) and above.