HiThink-Tech/Financial-API

HiThink‑Tech / Financial‑API – 同花顺金融数据服务

What it is – A commercial‑grade data platform, maintained by the Chinese market‑data provider 同花顺 (10 Juqka). It offers a unified, API‑key‑protected service for A‑share (China‑main‑board) market data, aimed at AI agents, quantitative researchers and developers.

Core capabilities

• Real‑time snapshots, collection‑auction snapshots and daily K‑line (OHLCV) for any A‑share ticker.
• Full‑company financial statements (income, balance‑sheet, cash‑flow) and a suite of valuation ratios (PE, PB, PS, PCF).
• Metadata: ticker lookup, trading‑calendar, corporate actions, split/bonus factors and forward‑/backward‑adjusted prices.
• Market‑wide data: index/sector constituents, fund/manager information, public futures & options contracts, and specialty lists such as limit‑up/down pools, “炸板” (break‑through) pools, hot‑stock and龙虎榜 rankings.
• Bulk market‑dump download for back‑testing or factor research.
• Local DuckDB‑based marketdb that can incrementally sync the whole market and be queried with plain SQL.

How to access

Method	Typical use‑case	Entry point
REST API	Embed in web services, apps, any language	https://fuyao.aicubes.cn/api/... (curl example in README)
MCP (Model‑Center‑Protocol)	Plug into LLM‑enabled tools like Claude Desktop, Cursor, Windsurf	Six hosted endpoints (a‑share, a‑share‑index, meta, fund, futures, options)
CLI (@hithink-tech/hithink-finance-cli)	Quick terminal queries, bulk download, data‑dump, local DuckDB init	hithink-finance … commands
Python SDK (python/ package)	Notebook exploration, custom scripts, integration with pandas/NumPy	python -m pip install -e ./python then use toolkit/fuyao scripts
Agent Skill (skills/hithink-finance/)	Let an AI agent automatically pick the best backend (API, MCP, CLI, SDK)	Installed via npx skills add …

Typical workflow

1. Obtain a single API key from the web portal and store it securely (HITHINK_FINANCE_API_KEY).
2. Choose the most convenient entry point (Skill → Agent, CLI for ad‑hoc, SDK for notebooks, REST for services).
3. Resolve a ticker to a unique thscode (e.g., 600519.SH).
4. Call the desired endpoint (snapshot, K‑line, financials, etc.).
5. For large results the system writes files locally (DuckDB or CSV) and returns only a path/summary to the caller.

Who should use it

• AI agents that need up‑to‑date market data (Claude, Cursor, etc.) – install the Skill so the agent can switch between API, MCP or CLI automatically.
• Quant researchers building back‑tests, factor models, or dashboards – use the CLI or marketdb to keep a local DuckDB copy and query with SQL.
• Developers building trading‑oriented web or mobile apps – call the REST API directly with any language.
• Data engineers needing bulk market dumps for data‑lake ingestion.

Safety & compliance notes

• The API key must never be hard‑coded or committed to source control.
• Large result sets must be written to disk; the service will not stream massive JSON payloads to avoid context overflow.
• The service provides no investment advice – callers must label any analysis as “non‑advisory”.
• Rate‑limit is soft (no hard quota) but users are asked to avoid bursty high‑concurrency traffic.

Project status

• Actively maintained monorepo (Python + Node.js CLI + Skill).
• Recent releases added collection‑auction data, expanded fund coverage, and introduced the unified Skill.
• Future roadmap mentions integration with the upcoming 同花顺 AI client for tighter LLM workflows.

License – MIT.

---

All information above is taken directly from the repository’s README; no additional features have been inferred.