Jquants mcp usage

jquants-mcp Usage Guide

jquants-mcp is an MCP server that retrieves Japanese stock market data via the J-Quants API. It serves data from a local SQLite cache (fast, offline) and falls back to the live API when needed.

Session Start

Always call health_check at the beginning of a session:

health_check()

• cache_ready: true means the cache has finished loading and screener tools are available.
• cache_ready: false (integrity is "pending") at first call after server start — wait 10–60 s and retry.
• today_cache_ready: true means today's equities data is already in cache.
• Check plan to know which endpoints and date ranges are available.

Daily Workflows

Prefer the composite one-call tools below over hand-assembling the same picture from primitive tools — they encode the correct definitions and are cached.

Morning market brief — one call

health_check()                           # read latest_cache_date from the response
get_market_briefing(date="2026-07-14")   # date is required (YYYY-MM-DD or YYYYMMDD)

Pass the latest cached trading day as date (health_check reports it as latest_cache_date) — a weekend/future date beyond the cache returns CacheNotReady.

Sections and how to read them:

• summary — advances/declines, 25-day advance/decline ratio, TOPIX change, and market margin ratio (margin figures need the Standard+ margin cache; null otherwise).
• sectors.top / sectors.bottom and sector_short_ratios — TSE 33-sector performance and short-sale pressure (short-sale ratios need the Standard+ short-ratio cache; empty/null otherwise).
• top_movers_up / top_movers_down / top_turnover_value — daily leaders.
• highlights — YTD high-low counts (ytd_new_highs / ytd_new_lows), volume surges, price-limit counts, and notable_stocks (RSI14 overbought/oversold from the 52w-high/low + price-limit universe).
• value_screen — the 年安・割安・高配当・好決算 screen (top n). null means the section was unavailable (e.g. cold financials cache); {"count": 0} means it ran and genuinely nothing matched. For different thresholds call get_value_stock_screen directly.
• trend_signals — distribution-day count and follow-through-day status (auto-detected rally start), so you rarely need to call the two detect tools yourself.

Value screening (年安・割安・高配当・好決算)

get_value_stock_screen()   # defaults: near_low_pct=5, max_per=15, max_pbr=1,
                           # min_yield=3.5, require_profit_increase=True

All criteria are ANDed; REITs are excluded. Definition pitfalls that ad-hoc screening gets wrong — this tool enforces the correct ones:

• 年安 (year-low) means the close is within near_low_pct % of the 52-week LOW (or a fresh 52w low was touched that day). "Down by half from the 52-week high" is NOT year-low — a stock 30% above its low never qualifies at the default.
• 好決算 (strong earnings) means the company-forecast net profit exceeds the latest FY actual (増益予想). Trailing black ink alone does not qualify.
• 高配当 uses the forward dividend forecast (same logic as get_dividend_yield_ranking), never the trailing dividend.

Empty result? Loosen one axis at a time: raise near_low_pct (e.g. 10–20), raise max_pbr, lower min_yield, or set require_profit_increase=False (items then still carry net_profit / forecast_net_profit so you can judge). market ("prime" / "standard" / "growth" / "tokyo_pro") and sector (S33 code) narrow the universe.

Each item also carries margin_ratio (信用倍率 = long/short margin balance) + margin_date for supply-demand context (Standard+; null otherwise) — no per-candidate get_stock_briefing call needed to judge crowdedness.

Single-stock deep dive

search_equities(name="住友商事")        # company name → code, when the code is unknown
get_stock_briefing(code="80530")        # price + FY financials + PER/PBR/yield in one call
                                        # (+ margin fields with the Standard+ cache)
get_fins_summary(code="80530")          # quarterly history and forecasts
get_technical_indicators(code="80530", date_from="2026-06-15")
                                        # SMA 5/25 + bb20 + RSI14 by default (date or
                                        # date_from/date_to required); add sma75 via
                                        # indicators=[...]; API fallback on cache miss
get_candlestick_data(code="80530")      # OHLCV + indicators as JSON for chart rendering

Sector value hunt

get_sector_briefing()                          # median PER/PBR/ROE per TSE 33 sector, cheapest first
get_valuation_ranking(sector="2050")           # then drill into one sector by PER (or metric="pbr")
get_dividend_yield_ranking(min_yield=4.0)      # forward-yield ranking (Kabutan-equivalent default)

Market temperature checks

• detect_price_change / get_advance_decline_ratio — breadth; ratio >120 overbought, <70 oversold.
• detect_distribution_days / detect_follow_through_day — TOPIX-based IBD-style trend signals (already embedded in the briefing's trend_signals).
• Leaders and sector moves beyond the briefing's top-n: get_top_movers, get_top_volume, get_top_turnover_value, get_sector_performance (sector_type="s17" for 17 sectors).

Earnings week

get_earnings_this_week()            # upcoming schedule, grouped by day
get_earnings_results_this_week()    # actual reported numbers with forecast progress
                                    # (empty on Free — results carry the 12-week delay)
get_equities_earnings_calendar(code="72030")   # next earnings date for one stock
                                    # (all plans; Light+ covers all issues incl. REITs,
                                    # Free is limited to March/September fiscal-year issues)

Plan Tiers and Date Ranges

Plan	Historical depth	Adds (over the tier below)
free	Latest 2 years (12-week delay)	—
light	Latest 5 years	investor_types, indices_bars_daily_topix, bulk downloads, minute bars (with tick add-on)
standard	Latest 10 years	margin_interest, margin_alert, short_ratio, short_sale_report, indices_bars_daily, derivatives options_225
premium	Full history	breakdown, derivatives futures/options, fins_details, fins_dividend, morning-session bars

Queries outside the plan's date range return an error — do not retry with the same date.

Stock Code Format

• Use 5-digit numeric codes: 72030 (Toyota), 86970 (Japan Exchange Group)
• 4-char alphanumeric codes (130A) are accepted and normalised to 130A0
• Do not use ticker symbols or company names as codes (use search_equities for name lookup)

Tool Categories

Composite briefings (cache-first, start here)

• get_market_briefing — whole-market daily brief (see workflow above)
• get_sector_briefing — sector median PER/PBR/ROE
• get_stock_briefing — one-page single-stock brief
• get_value_stock_screen — combined value screen (年安・割安・高配当・好決算)

Market overview & rankings (cache-only)

• detect_price_change, get_advance_decline_ratio, get_top_movers, get_top_volume, get_top_turnover_value, get_sector_performance
• get_dividend_yield_ranking, get_valuation_ranking

Equities

• get_equities_master — listed company master (name, sector, market)
• search_equities — company name → code reverse lookup (cache-only)
• get_equities_bars_daily — OHLCV daily bars with split adjustment
• get_equities_bars_minute — intraday minute bars (light+ with tick add-on)
• get_equities_bars_daily_am — today's morning session bars (premium)
• get_equities_investor_types — weekly investor type flows (light+)
• get_equities_earnings_calendar — scheduled earnings dates (all plans; Light+ covers all issues incl. REITs, Free is limited to March/September fiscal-year issues)
• get_earnings_this_week / get_earnings_results_this_week — weekly earnings schedule / results (cache-only)

Financials

• get_fins_summary — quarterly earnings summary (EPS, revenue, guidance)
• get_fins_details — detailed P&L / balance sheet items (premium)
• get_fins_dividend — dividend history (premium)

Markets

• get_markets_margin_interest / get_markets_margin_alert — margin trading data (standard+)
• get_markets_short_ratio / get_markets_short_sale_report — short-selling data (standard+)
• get_markets_breakdown — trading breakdown by investor type (premium)
• get_markets_calendar — trading calendar (holidays, market open/close)

Indices

• get_indices_bars_daily_topix — TOPIX daily bars (light+)
• get_indices_bars_daily — other indices (standard+)

Screener (fast)

The 52w/YTD detectors and detect_consecutive_dividend_increase hit a nightly pre-built cache (~0.01 s on default params); the other screeners compute from the cached daily bars. The 52w/YTD detectors (and their _range variants) are limited to the last 52 weeks — older dates return OutOfCacheRange immediately; do not retry. The other screeners accept any date within your plan's window.

• detect_52w_high_low / detect_ytd_high_low — fresh high/low breakouts on a date. These list stocks that SET a new high/low that day. There is no plain near-the-low lister: get_value_stock_screen finds near-low stocks that ALSO pass its value criteria (loosen them per the workflow section — note it always excludes REITs and net-loss / negative-book stocks).
• detect_52w_high_low_range / detect_ytd_high_low_range — multi-day variants; prefer one _range call over per-day loops: detect_52w_high_low_range(date_from="2026-07-06", date_to="2026-07-10")
• detect_price_limit — stocks at daily price limit (ストップ高/安)
• detect_volume_surge — unusual volume spikes vs the 20-day average
• compare_close_vs_vwap — close above/below VWAP (per-code, always returns data)
• detect_consecutive_dividend_increase — consecutive annual dividend increases (連続増配)
• detect_distribution_days / detect_follow_through_day — TOPIX trend signals

Most screeners are cache-only, but compare_close_vs_vwap and per-code detect_volume_surge calls fall back to the live API on a cache miss (they need a working API key there), as does get_technical_indicators.

detect_price_limit, detect_volume_surge, and the 52w/YTD detectors (+_range) accept detail: bool = False (detect_consecutive_dividend_increase always returns full data):

• detail=False (default): summary counts only — prefer this on mobile and whenever you only need totals.
• detail=True: full per-stock data array.

A code= argument bypasses the pre-built cross-sectional payload (so newly-listed stocks are not dropped) — the tool then computes from cached daily bars, with the API fallbacks noted above where applicable.

Charts

• get_candlestick_data — OHLCV + indicator data as JSON for React artifact rendering
• get_comparison_chart_data — multi-stock comparison; up to 10 codes, mode="return_pct" (default) or "price"

Bulk (light+)

• get_bulk_list — list available bulk data files
• get_bulk_download_url — get a signed URL for bulk file download

Derivatives

• get_derivatives_bars_daily_futures / get_derivatives_bars_daily_options — premium
• get_derivatives_bars_daily_options_225 — Nikkei 225 options (standard+)

Cache Management

cache_status

cache_status()

Shows row counts per table, file size, and detected plan. Does not return market data.

cache_clear — USE WITH CARE

Call	Effect
cache_clear(table="response_cache")	Clears only the API response cache (Tier 2). Safe — historical Tier 1 data is preserved. Use when you suspect stale live-API responses.
cache_clear() (no argument)	Clears ALL data including all Tier 1 historical rows. Only use if you intend a full cache rebuild.

Never call bare cache_clear() unless you are prepared to re-run the full bulk fetch (can take ~1 hour).

Multi-user Mode (Cloud Run)

1. Authenticate via Google OAuth in the MCP client
2. Register your J-Quants API key at https://<host>/settings (recommended) or via register_api_key (delete_api_key removes it); the server auto-detects your plan and applies the correct date-range restrictions per user

Single-user local mode instead uses the JQUANTS_API_KEY env var or the [jquants] section in the config file.