wind-mcp 1.0.0

MCP server for Wind Financial Terminal (万得)

Wind MCP

A Model Context Protocol server that gives AI assistants direct access to Wind Financial Terminal data.

---

Wind MCP bridges the Wind Financial Terminal (万得) with AI assistants via the Model Context Protocol. It exposes 19 tools covering market data, fundamentals, macro economics, screening, estimates, and date utilities — all accessible through natural language.

Companion project to Bloomberg-MCP. Same architecture, same philosophy — but for Wind (万得).

You: "拉一下贵州茅台过去20天的量价数据"

Claude: runs wind_historical with codes=600519.SH, fields=close,volume
        → Returns structured price/volume time series

You: "Find me A-share stocks with PE < 15 and ROE > 20%"

Claude: runs wind_dynamic_screen with universe=A-shares, filters, rank
        → Returns filtered and ranked stock list

Architecture

graph TB
    subgraph Clients
        CC[Claude Code]
        WC[Web Client]
        CA[Custom App]
    end

    subgraph "Wind MCP Server"
        direction TB
        MCP["FastMCP Server<br/><i>19 tools exposed</i>"]

        subgraph Handlers["Handler Layer"]
            direction LR
            SNAP[Snapshot & Historical]
            INTRA[Minute Bars & Ticks]
            MACRO[Macro & Sector]
            SCREEN[Screening]
            EST[Estimates & Holders]
            CAL[Calendar & Dates]
        end

        subgraph Core["Core Layer"]
            direction LR
            SESSION["WindSession<br/><i>Singleton + atexit</i>"]
            CACHE["WindCache<br/><i>TTL-based</i>"]
            PARSER["Parser<br/><i>12 parsers</i>"]
            CONV["Converter<br/><i>BBG → Wind</i>"]
        end
    end

    WIND["Wind Terminal<br/><i>WindPy (w.start)</i>"]

    CC -- stdio --> MCP
    WC -- HTTP/SSE --> MCP
    CA -- HTTP/SSE --> MCP
    MCP --> Handlers
    Handlers --> Core
    SESSION <--> WIND

    style MCP fill:#1a73e8,stroke:#1557b0,color:#fff
    style WIND fill:#e65100,stroke:#bf360c,color:#fff
    style SESSION fill:#2e7d32,stroke:#1b5e20,color:#fff
    style CACHE fill:#2e7d32,stroke:#1b5e20,color:#fff

Tools Overview (19 tools)

graph LR
    subgraph "Market Data (5)"
        T1["wind_snapshot<br/><i>WSS current data</i>"]
        T2["wind_historical<br/><i>WSD time series</i>"]
        T3["wind_minute_bars<br/><i>WSI candles</i>"]
        T4["wind_ticks<br/><i>WST raw ticks</i>"]
        T5["wind_realtime<br/><i>WSQ live quotes</i>"]
    end

    subgraph "Fundamentals & Screening (5)"
        T6["wind_dataset<br/><i>WSET tables</i>"]
        T7["wind_screen<br/><i>WEQS screens</i>"]
        T8["wind_dynamic_screen<br/><i>Custom filter + rank</i>"]
        T9["wind_estimates<br/><i>Consensus</i>"]
        T10["wind_holders<br/><i>Shareholding</i>"]
    end

    subgraph "Macro & Sector (3)"
        T11["wind_macro<br/><i>EDB series</i>"]
        T12["wind_sector_series<br/><i>WSES time series</i>"]
        T13["wind_sector_snapshot<br/><i>WSEE snapshot</i>"]
    end

    subgraph "Connect & Calendar (3)"
        T14["wind_stock_connect<br/><i>Northbound flows</i>"]
        T15["wind_event_calendar<br/><i>IPO, earnings...</i>"]
        T16["wind_list_fieldsets<br/><i>Available FieldSets</i>"]
    end

    subgraph "Date Utilities (3)"
        T17["wind_trading_days<br/><i>TDAYS</i>"]
        T18["wind_date_offset<br/><i>TDAYSOFFSET</i>"]
        T19["wind_days_count<br/><i>TDAYSCOUNT</i>"]
    end

    style T6 fill:#1a73e8,stroke:#1557b0,color:#fff
    style T7 fill:#1a73e8,stroke:#1557b0,color:#fff
    style T8 fill:#1a73e8,stroke:#1557b0,color:#fff
    style T11 fill:#7b1fa2,stroke:#6a1b9a,color:#fff
    style T12 fill:#7b1fa2,stroke:#6a1b9a,color:#fff
    style T13 fill:#7b1fa2,stroke:#6a1b9a,color:#fff

Key Features

• Bloomberg → Wind Ticker Converter — Accepts Bloomberg-style identifiers (AAPL US Equity, 700 HK Equity, 601012 CH Equity) and auto-converts to Wind codes (AAPL.O, 00700.HK, 601012.SH). Covers US, HK, JP, LN, CH equities, indices, commodities, and currencies.
• Modular architecture — server.py is a thin entry point. Handlers, models, formatters, parsers, and converters cleanly separated.
• 16 FieldSet shortcuts — Pre-defined field collections (PRICE, VALUATION, PROFITABILITY, etc.) that expand to Wind field mnemonics.
• Cache layer — TTL-based in-memory cache with data-type-aware expiration (30s for realtime, 24h for static).
• 12 WindPy parsers — Handle WSS, WSD, WSI, WST, WSQ, WSET, EDB, WSES, WSEE, TDAYS, TDAYSOFFSET, TDAYSCOUNT with NaN→None cleanup and column-to-row transposition.
• Dynamic Screening — Custom filtering with operators (gt, lt, between, in, eq) + ranking + slicing, without knowing Wind mnemonics.

Data Flow

sequenceDiagram
    participant Client as AI Assistant
    participant MCP as MCP Server
    participant Conv as Ticker Converter
    participant Expand as Field Expander
    participant Cache as Cache Layer
    participant Session as WindSession
    participant Wind as Wind Terminal

    Client->>MCP: Tool call (JSON)
    MCP->>Conv: Convert BBG → Wind codes
    Conv-->>MCP: Wind codes

    alt FieldSet shortcuts used
        MCP->>Expand: Expand FieldSet shortcuts
        Expand-->>MCP: Resolved field list
    end

    MCP->>Cache: Check cache
    alt Cache hit
        Cache-->>MCP: Cached result
    else Cache miss
        MCP->>Session: Call WindPy API
        Session->>Wind: w.wsd() / w.wss() / ...
        Wind-->>Session: WindData response
        Session-->>MCP: Parsed dicts
        MCP->>Cache: Store with TTL
    end

    alt Markdown format
        MCP-->>Client: Formatted table
    else JSON format
        MCP-->>Client: Structured JSON
    end

Tool Reference

Market Data

Tool	WindPy API	Description	Key Parameters
wind_snapshot	WSS	Current field values for any security	codes, fields, trade_date
wind_historical	WSD	Time series with configurable periodicity	codes, fields, begin_date, end_date
wind_minute_bars	WSI	OHLCV intraday candles (1/3/5/15/30/60 min)	codes, fields, begin_time, end_time, bar_size
wind_ticks	WST	Raw tick-level trade data	codes, fields, begin_time, end_time
wind_realtime	WSQ	Live streaming quotes	codes, fields

Fundamentals & Screening

Tool	WindPy API	Description	Key Parameters
wind_dataset	WSET	Structured datasets (index members, IPOs, etc.)	table_name, options
wind_screen	WEQS	Execute saved Wind stock screens	screen_name, options
wind_dynamic_screen	WSS	Custom filter + rank + slice	universe, fields, filters, rank_by, top_n
wind_estimates	WSS	Consensus estimates and target prices	codes, metrics, year
wind_holders	WSS/WSET	Top holders, institutional, fund holdings	codes, holder_type, date

Macro & Sector

Tool	WindPy API	Description	Key Parameters
wind_macro	EDB	Macroeconomic indicator time series	codes, begin_date, end_date
wind_sector_series	WSES	Sector-level time series	codes, fields, begin_date, end_date
wind_sector_snapshot	WSEE	Sector-level current snapshot	codes, fields, trade_date

Stock Connect & Calendar

Tool	WindPy API	Description	Key Parameters
wind_stock_connect	WSET/WSS	Northbound/Southbound flow data	codes, direction, date
wind_event_calendar	WSET	IPO, earnings, dividend calendars	event_type, begin_date, end_date
wind_list_fieldsets	—	List all available FieldSet shortcuts	—

Date Utilities

Tool	WindPy API	Description	Key Parameters
wind_trading_days	TDAYS	List trading days in a range	begin_date, end_date, calendar
wind_date_offset	TDAYSOFFSET	Offset a date by N trading days	date, offset, calendar
wind_days_count	TDAYSCOUNT	Count trading days between two dates	begin_date, end_date, calendar

All tools support response_format: "markdown" (default) or "json".

FieldSet Shortcuts

Instead of remembering Wind field mnemonics, use shorthand names that expand to multiple fields.

FieldSet	Fields	Key Wind Mnemonics
PRICE	5	close, open, high, low, pct_chg
MOMENTUM	4	pct_chg, pct_chg_5d, pct_chg_1m, pct_chg_ytd
VOLUME_PROFILE	4	volume, amt, turn, free_turn
VALUATION	5	pe_ttm, pb_lf, ps_ttm, pcf_ocf_ttm, ev2_to_ebitda
VALUATION_EXTENDED	9	+ pe_est, dividend_yield, total_mkt_cap, ev
PROFITABILITY	6	roe_ttm, roa_ttm, grossprofitmargin, netprofitmargin, operatingprofitmargin, roic
GROWTH	4	yoyprofit, yoyrevenue, yoyocf, qfa_yoygr
BALANCE_SHEET	6	debttoassets, current_ratio, quick_ratio, cashflow_to_debt, longdebttodebt, equity_ratio
CASH_FLOW	5	ocfps, fcf, cf_from_ops, capex, dividendps
TECHNICAL	5	rsi, macd, vol_20d, beta_100w, atr_14d
ANALYST	4	rating_avg, est_target_price, est_eps_fy1, est_net_profit_fy1
ESTIMATE_REVISIONS	4	est_eps_chg_4w, est_eps_chg_13w, est_num_up, est_num_down
SECTOR	2	industry_sw, industry_sw_lv2
NORTHBOUND	4	sh_hk_share_pct, sh_hk_share_chg, sh_hk_share_amt, sh_hk_share_rank
MARGIN	3	margin_buy_bal, margin_sell_bal, margin_net_bal
RISK	5	beta_100w, volatility_20d, volatility_60d, sharpe_20d, max_drawdown_1y
SCREENING_FULL	50+	All of the above combined

Bloomberg → Wind Ticker Conversion

Wind MCP automatically converts Bloomberg-style identifiers to Wind codes:

Bloomberg Format	Wind Code	Market
AAPL US Equity	AAPL.O	US (NASDAQ)
JPM US Equity	JPM.N	US (NYSE)
700 HK Equity	00700.HK	Hong Kong
7203 JP Equity	7203.T	Japan
VOD LN Equity	VOD.L	London
601012 CH Equity	601012.SH	A-share (Shanghai)
000001 CH Equity	000001.SZ	A-share (Shenzhen)
300750 CH Equity	300750.SZ	A-share (ChiNext)
SPX Index	SPX.GI	Index
HSI Index	HSI.HI	Index
CL1 Comdty	CL.NYM	Commodity
EURUSD Curncy	EURUSD.FX	Currency
600519.SH	600519.SH	Passthrough (already Wind)

Dynamic Screening

Build custom screens with pre-validated field sets, filters, and ranking — no need to know Wind field mnemonics.

flowchart LR
    A["Universe<br/><i>index, sector,<br/>or ticker list</i>"] --> B["Field Expansion<br/><i>FieldSet shortcuts<br/>→ Wind fields</i>"]
    B --> C["Wind API<br/><i>WSS request</i>"]
    C --> D["Filter<br/><i>gt, lt, between,<br/>in, eq, ...</i>"]
    D --> E["Rank & Slice<br/><i>rank_by + top_n</i>"]
    E --> F["Response<br/><i>Markdown table<br/>or JSON</i>"]

    style A fill:#e8f5e9,stroke:#2e7d32
    style C fill:#fff3e0,stroke:#ff6f00
    style F fill:#e3f2fd,stroke:#1a73e8

Filter Operators

Operator	Description	Example
gt / gte	Greater than (or equal)	{"field": "roe_ttm", "op": "gt", "value": 20}
lt / lte	Less than (or equal)	{"field": "pe_ttm", "op": "lt", "value": 15}
eq / neq	Equals / not equals	{"field": "industry_sw", "op": "eq", "value": "电子"}
between	Range (inclusive)	{"field": "pe_ttm", "op": "between", "value": [10, 25]}
in	Value in list	{"field": "industry_sw", "op": "in", "value": ["电子", "计算机"]}

Example: Find Undervalued High-ROE A-shares

{
  "universe": "sector:全部A股",
  "fields": ["PRICE", "VALUATION", "PROFITABILITY"],
  "filters": [
    {"field": "pe_ttm", "op": "lt", "value": 15},
    {"field": "roe_ttm", "op": "gt", "value": 20}
  ],
  "rank_by": "roe_ttm",
  "rank_descending": true,
  "top_n": 20
}

Cache Layer

Built-in cache reduces Wind API load with data-type-aware TTLs:

Data Type	TTL	Rationale
Realtime quotes (WSQ)	30 seconds	Near real-time
Historical (WSD)	12 hours	End-of-day data stable
Snapshot (WSS)	5 minutes	Moderate refresh
Static reference	24 hours	Rarely changes
Macro data (EDB)	6 hours	Periodic updates
Dataset (WSET)	1 hour	Moderate refresh

Project Structure

wind-mcp/
├── pyproject.toml
├── README.md
├── LICENSE
├── CLAUDE.md
├── run_server.bat / run_server.ps1
├── src/wind_mcp/
│   ├── __init__.py
│   ├── __main__.py
│   ├── server.py                  # FastMCP entry point, 19 tools
│   ├── formatters.py              # Markdown table & JSON output
│   ├── utils.py                   # Date normalization, validation
│   ├── core/
│   │   ├── session.py             # WindSession singleton + atexit
│   │   ├── cache.py               # TTL cache with hit/miss stats
│   │   ├── parser.py              # 12 WindData parsers
│   │   └── converter.py           # Bloomberg → Wind ticker converter
│   ├── models/
│   │   ├── enums.py               # ResponseFormat, Periodicity, etc.
│   │   └── inputs.py              # 18 Pydantic input models
│   ├── tools/
│   │   ├── fieldsets.py           # 16 FieldSet definitions
│   │   └── field_expander.py      # FieldSet → field list resolver
│   ├── handlers/                  # 12 handler modules
│   │   ├── snapshot.py            # WSS
│   │   ├── historical.py          # WSD
│   │   ├── minute_bars.py         # WSI
│   │   ├── ticks.py               # WST
│   │   ├── realtime.py            # WSQ
│   │   ├── dataset.py             # WSET
│   │   ├── macro.py               # EDB
│   │   ├── sector.py              # WSES / WSEE
│   │   ├── screening.py           # WEQS + dynamic screen
│   │   ├── estimates.py           # Consensus estimates
│   │   ├── holders.py             # Holder analysis
│   │   ├── stock_connect.py       # Northbound/Southbound
│   │   ├── calendar.py            # Event calendar
│   │   └── dates.py               # TDAYS / TDAYSOFFSET / TDAYSCOUNT
│   └── data/                      # Static mapping files
│       ├── us_exchange_map.json   # ~200 US tickers → exchange
│       ├── index_map.json         # BBG → Wind index mapping
│       ├── commodity_map.json     # BBG → Wind commodity mapping
│       └── currency_map.json      # BBG → Wind currency mapping
├── tests/                         # 42 unit tests (no Wind needed)
│   ├── test_parser.py
│   ├── test_fieldsets.py
│   ├── test_cache.py
│   ├── test_formatters.py
│   └── test_converter.py
└── examples/
    ├── basic_usage.py
    └── screening_example.py

Installation

Prerequisites

• Python 3.10+
• Wind Financial Terminal (万得) running and logged in
• WindPy installed (pip install WindPy or repair via Wind Terminal)

Setup

# Clone the repository
git clone https://github.com/QmQsun/Wind-MCP.git
cd Wind-MCP

# Install
pip install .            # standard install
# or: pip install -e .   # editable mode (for development)

Configure Claude Code

Add to your Claude Code MCP settings (.mcp.json):

{
  "mcpServers": {
    "wind-mcp": {
      "command": "python",
      "args": ["-m", "wind_mcp.server"],
      "cwd": "/path/to/Wind-MCP/src"
    }
  }
}

Install from PyPI

pip install wind-mcp

Quick Start

As an MCP Server

# stdio (default — for Claude Code)
python -m wind_mcp.server

# HTTP transport (for web clients)
python -m wind_mcp.server --transport http --port 8080

# SSE transport (for streaming clients)
python -m wind_mcp.server --transport sse --port 8080

Windows

# Command Prompt
run_server.bat

# PowerShell
.\run_server.ps1

Wind API Coverage

graph LR
    subgraph "Wind Terminal (WindPy)"
        WSS["WSS<br/><i>Snapshot</i>"]
        WSD["WSD<br/><i>Historical</i>"]
        WSI["WSI<br/><i>Minute bars</i>"]
        WST["WST<br/><i>Tick data</i>"]
        WSQ["WSQ<br/><i>Realtime</i>"]
        WSET["WSET<br/><i>Datasets</i>"]
        EDB["EDB<br/><i>Macro</i>"]
        WSES["WSES<br/><i>Sector series</i>"]
        WSEE["WSEE<br/><i>Sector snapshot</i>"]
        WEQS["WEQS<br/><i>Stock screener</i>"]
        TDAYS["TDAYS<br/><i>Trading days</i>"]
    end

    WSS --- T1[Snapshot + Estimates + Holders + Screen]
    WSD --- T2[Historical Time Series]
    WSI --- T3[Intraday Bars]
    WST --- T4[Tick Data]
    WSQ --- T5[Realtime Quotes]
    WSET --- T6[Datasets + Connect + Calendar]
    EDB --- T7[Macro Indicators]
    WSES --- T8[Sector Time Series]
    WSEE --- T9[Sector Snapshot]
    WEQS --- T10[Saved Screens]
    TDAYS --- T11[Date Utilities]

    style WSS fill:#1a73e8,stroke:#1557b0,color:#fff
    style WSD fill:#1a73e8,stroke:#1557b0,color:#fff
    style EDB fill:#7b1fa2,stroke:#6a1b9a,color:#fff
    style WEQS fill:#e65100,stroke:#bf360c,color:#fff

Contributing

Contributions welcome! Please open an issue or submit a pull request.

pip install -e ".[dev]"
pytest                    # Unit tests (no Wind Terminal needed)
black src/ tests/
ruff check src/ tests/

Contributors

	Contributor	Role
1	QmQsun	Architecture, spec design, code review
2	Claude (Anthropic)	Implementation, code generation, testing

Related Projects

• Bloomberg-MCP — Same architecture for Bloomberg Terminal

License

MIT — see LICENSE for details.