Find the Right Free Financial Database
for Research and AI如何选择适合研究与 AI 的
免费金融数据库
Compare free sources for prices, fundamentals, and filings—then check coverage, limits, licensing, and API access before you build.
比较价格、基本面与监管申报数据源，
并在构建前核对覆盖范围、限制、许可和 API 接入。
TL;DR摘要
A free financial database provides market or company data at no cost, usually with limits on coverage, history, calls, or commercial use.
Official filings, public economic datasets, exchange resources, and freemium APIs cover different parts of the research stack.
“Free” rarely means unlimited. Data quality, licensing, update speed, and stable access matter more than row count alone.
Use a capability layer when an agent must discover the right source, inspect its contract, and call it with an auditable trail.
免费金融数据库以零费用提供市场或公司数据，但通常会限制覆盖范围、历史深度、调用次数或商业用途。
监管申报文件、公共经济数据、交易所资源和提供免费套餐的 API，分别覆盖研究链路的不同环节。
“免费”很少等于无限使用。数据质量、许可、更新速度和稳定接入通常比数据量更重要。
当 Agent 需要查找数据源、检查接口约束并保留可审计的调用记录时，可以使用 QVeris 统一管理调用流程。
Free financial database, dataset, or API?免费金融数据库、数据集和 API 有什么区别
People use “financial database” to describe very different products: a website where they can search records, a downloadable dataset, an API that returns current values, or a database they operate themselves. Choose the form by the task rather than expecting one free service to satisfy every need.
大家口中的“金融数据库”可能指完全不同的东西：能检索记录的网站、可下载的数据集、返回最新数值的 API，或自己维护的数据仓库。应先从任务出发选择形式，而不是期待一个免费服务包办全部需求。
| Form形式 | Best when适合情况 | What you control你能控制什么 | Main limitation主要限制 |
|---|---|---|---|
| Searchable portal在线检索门户 | A person needs to look up a filing, series or company occasionally.人工偶尔查询申报文件、经济序列或公司信息。 | Filters, exports and citations offered by the publisher.发布方提供的筛选、导出与引用功能。 | Hard to automate or reproduce at scale.难以规模化自动执行，也不便复现实验。 |
| Downloadable dataset可下载数据集 | You need a fixed snapshot for research, teaching or a one-off model.研究、教学或一次性模型需要固定快照。 | Local validation, versioning and transformations.本地验证、版本管理和转换过程。 | The snapshot becomes stale unless you refresh it.如果不主动更新，快照会逐渐过期。 |
| API | An app needs parameterized or frequently refreshed data.应用需要按参数查询，或频繁获取更新数据。 | Request timing, cache and your internal response model.请求时机、缓存策略和内部响应模型。 | Quotas, network failures and provider schema changes.调用额度、网络故障和供应商结构变化。 |
| Self-managed database自建数据库 | You join sources, query repeatedly, preserve history or serve several applications.需要合并多个来源、反复查询、保存历史，或同时服务多个应用。 | Schema, quality rules, point-in-time history, indexes and access.数据结构、质量规则、历史时点、索引和访问权限。 | You own ingestion, monitoring, licensing records and maintenance.采集、监控、许可记录和维护都由自己承担。 |
Useful rule: start with a portal for discovery, preserve a dataset snapshot for reproducibility, use APIs for incremental refreshes, and build your own database only when repeated joins, history or product access justify the maintenance.
实用顺序：先用门户发现数据，再保留快照保证研究可复现；需要持续更新时接 API；只有当反复关联查询、历史保存或产品访问确实需要时，再建设自己的数据库。
What a free financial database can cover免费金融数据库可以覆盖哪些数据
No single free source reliably covers every asset, field, geography, and time horizon. Define the decision you need to support, then choose the smallest combination of sources that supplies the required data on acceptable terms.
没有一个免费数据源能可靠覆盖所有资产、字段、地区和时间范围。先明确要支持的研究或产品决策，再选择满足数据与许可要求的最小数据源组合。
Daily or intraday prices, volume, splits, dividends, symbols, exchanges, currencies, and trading calendars form the base for screening and charting.
Income statements, balance sheets, cash flow, ratios, shares outstanding, and corporate actions support valuation and financial analysis.
Regulatory filings and issuer announcements are often the most authoritative source for company-reported facts, but parsing them takes work.
Rates, inflation, labor, trade, and sector indicators provide context. News, sentiment, and alternative datasets usually carry stricter terms.
日线或盘中价格、成交量、拆股、分红、证券代码、交易所、币种和交易日历，是筛选和图表分析的基础。
利润表、资产负债表、现金流量表、财务比率、已发行在外股份和公司行为，可用于估值和财务分析。
监管申报文件和发行人公告通常是公司披露事实最权威的来源，但结构化解析与字段标准化需要额外工作。
利率、通胀、就业、贸易和行业指标提供宏观背景；新闻、市场情绪和另类数据通常有更严格的使用条款。
Free financial database sources compared免费金融数据库来源对比
Prefer an original publisher for the facts it owns, then add a normalized API only where convenience is worth the extra dependency. The examples below are starting points, not interchangeable substitutes.
某项事实由谁直接发布，就应优先核对谁的原始数据；只有在便利性确实值得增加一层依赖时，再补充标准化 API。下面这些来源适合作为起点，但不能互相完全替代。
| Source来源 | Typical data常见数据 | Best for适合场景 | Watch for注意事项 |
|---|---|---|---|
| SEC EDGAR | US filings, submissions and XBRL company facts.美国公司申报、申报历史和 XBRL 财务事实。 | Primary-source company research and auditable fundamentals.一手公司研究与可追溯基本面。 | Taxonomy, units, filing periods, amendments and issuer calendars.分类标准、单位、申报期间、更正文件和公司财年。 |
| FRED / ALFRED | Rates, inflation, labor, GDP, financial conditions and vintages.利率、通胀、就业、GDP、金融条件和历史版本。 | Macroeconomic research and point-in-time release analysis.宏观研究，以及按历史发布日期还原当时可得数据。 | Series-specific units, revisions, frequencies and upstream rights.各序列的单位、修订方式、频率与上游许可。 |
| World Bank | Cross-country development, macro and financial indicators.跨国家的发展、宏观与金融指标。 | Country panels and long-run comparisons.国家面板与长期比较。 | Indicator metadata, source notes, missing years and revisions.指标元数据、来源说明、缺失年份和修订。 |
| Freemium market APIs提供免费套餐的行情 API | Prices, symbols, fundamentals and derived indicators.价格、证券代码、基本面与派生指标。 | Prototypes, dashboards and incremental refreshes.原型、仪表盘与增量更新。 | Quotas, delays, exchange rights, adjusted history and delisted coverage.配额、延迟、交易所数据权限、复权历史和退市证券覆盖。 |
Choose a storage architecture that matches the workload按实际查询负载选择金融数据存储架构
“Database” does not automatically mean a large cloud warehouse. A local analytical file can be enough for one researcher; a relational database is often better for shared identifiers and fundamentals; a warehouse or lake becomes useful when history, many sources and scheduled pipelines grow.
“数据库”并不等于一开始就上大型云仓库。单人研究可能只需要本地分析文件；多人共享标识符和基本面时，关系型数据库通常更合适；只有当历史规模、来源数量和定时任务不断增长时，数据仓库或数据湖才更有价值。
| Storage pattern存储方式 | Good fit适用场景 | Strength优势 | When to move on何时需要升级 |
|---|---|---|---|
| CSV / Parquet snapshotsCSV／Parquet 快照 | One analyst, immutable research inputs and batch notebooks.单人分析、不可变研究输入和批量 Notebook。 | Low setup cost, portable and easy to archive.搭建成本低，便于传递和归档。 | Concurrent edits, repeated joins or many small updates become painful.多人并发、反复关联或频繁小更新开始变得困难时。 |
| Embedded analytical database嵌入式分析数据库 | Local SQL across files and medium-size research datasets.在本地通过 SQL 查询多个文件和中等规模研究数据。 | Fast analytical queries without running a server.无需维护服务器，也能高效做分析查询。 | Many writers, central permissions or always-on application traffic appear.出现多写入方、集中权限控制或持续在线业务流量时。 |
| Relational database关系型数据库 | Shared symbols, entities, filings, fundamentals and application queries.多人共享的证券、实体、申报、基本面和应用查询。 | Constraints, transactions, indexes, permissions and dependable joins.约束、事务、索引、权限与可靠关联。 | Very large append-only histories or broad analytical scans dominate.不适合以超大规模追加式历史数据或大范围分析扫描为主要负载的场景。 |
| Warehouse / lakehouse数据仓库／湖仓 | Many sources, long histories, scheduled pipelines and team analytics.来源众多、历史很长、定时管道和团队分析。 | Scalable storage, partitioned scans and workload separation.可扩展存储、分区扫描和工作负载隔离。 | Do not start here unless scale and governance justify the cost.除非规模和治理需求已经明确，否则不必从这里起步。 |
A practical schema for a free financial database免费金融数据库应具备怎样的基础数据结构
Design around stable entities and events rather than one provider’s response. Keep identity, observations, source lineage and revisions separate so a ticker change or restatement does not overwrite history.
数据结构应围绕稳定实体和事件设计，而不是照抄某家供应商的响应。身份、观测值、来源记录和修订记录要分开保存，证券代码变化或财务重述才不会覆盖历史。
Give companies, legal entities, securities, listings and exchanges separate identifiers. A ticker belongs to a listing and can change; it should not be the permanent company key.
Store observation time, release time, ingestion time, value, unit, currency, frequency, adjustment status and source. Use a composite key that reflects the real grain.
Preserve accession or document ID, form, filed date, report period, fiscal period, taxonomy concept, unit and amendment status. A normalized metric should still link back to the filed fact.
Record request URL or file, retrieval time, checksum, adapter version and validation result. Revisions should create a new version instead of silently mutating the old row.
公司、法律实体、证券、上市地点和交易所应使用不同标识符。Ticker 属于某个上市证券，而且可能变化，不能直接当作永久公司主键。
保存观测时间、发布时间、入库时间、数值、单位、币种、频率、复权状态和来源，并按真实数据粒度设计复合主键。
保留 accession 或文档编号、表单类型、申报日期、报告期、财务期间、分类概念、单位和更正状态。标准化指标也必须能追溯到原始申报事实。
记录请求 URL 或文件、抓取时间、校验和、适配器版本和验证结果。数据发生修订时应生成新版本，而不是悄悄改掉旧行。
Point-in-time rule: if the database supports backtests or historical decisions, keep both the period a value describes and the time it became available. A clean final series without release or revision history can create look-ahead bias.
历史时点原则：如果数据库用于回测或还原历史决策，就必须同时保存数值对应的期间，以及它真正对外可见的时间。只有最终修订序列、没有发布与修订历史，会造成前视偏差。
How to choose a free financial database如何选择免费金融数据库
Write down assets, markets, fields, frequency, historical depth, and acceptable delay before comparing providers.
Confirm whether storage, derived data, public display, commercial use, model training, and redistribution are allowed.
Check splits, dividends, ticker changes, amended filings, missing values, time zones, and point-in-time behavior.
Cache responsibly, respect quotas, record provenance, and isolate provider fields behind your own schema.
Include parsing, storage, refresh jobs, monitoring, data reviews and license checks. A zero-dollar feed can still be expensive to make dependable.
Measure missing records, latency, schema changes, correction frequency, quota use and manual repair time before a production commitment.
在比较供应商前，明确资产、市场、字段、频率、历史深度和可接受延迟。
确认是否允许存储、衍生、公开展示、商业使用、模型训练和再分发。
检查拆股、分红、代码变更、更正公告、缺失值、时区和时点数据。
合理缓存、遵守配额、记录来源，并通过内部数据结构隔离供应商字段。
把解析、存储、刷新任务、监控、数据复核和许可审查都算进去。接口价格为零，并不代表把它做可靠也没有成本。
正式投入前，记录缺失数据、延迟、字段变化、更正频率、额度消耗和人工修复时间。
Use free financial data with QVeris通过 QVeris 使用免费金融数据
A database stores data; an agent still needs to find the right capability, inspect its inputs, call it, and preserve provenance. QVeris provides that capability-routing layer across financial APIs and tools.
数据库负责存储数据；Agent 仍需要查找所需能力、检查输入、执行调用并保留数据来源。QVeris 可以帮助 Agent 查找并调用相应的金融 API 和工具。
- Discover relevant financial capabilities instead of hardcoding one provider too early.
- Inspect schemas and requirements before an agent sends a live request.
- Keep calls and results traceable when research depends on external data.
- 先查找相关金融数据能力，避免过早把应用写死在单一供应商上。
- 在 Agent 发起真实请求前检查数据结构与调用要求。
- 当研究依赖外部数据时，保持调用与结果可追溯。
FAQ常见问题
Yes for specific datasets and research uses, but rarely as one unlimited source. Most free options restrict coverage, history, volume, or commercial use.
Regulatory filings are the primary source in many markets, though normalized fields and comparisons usually require parsing.
Only when the license and exchange rules allow it. Check display, storage, derived works, attribution, and redistribution separately.
Yes. Use a documented API contract, validate outputs, respect quotas, and retain timestamps and provenance.
特定数据集和研究用途有免费来源，但很少有一个来源无限覆盖全部需求。常见限制包括范围、历史、请求量与商业用途。
许多市场的监管申报文件是一手来源，权威性较高，但字段标准化和跨公司比较通常还需要解析。
只有在许可与交易所规则允许时才可以。应分别确认展示、存储、衍生、署名和再分发权利。
可以。应使用明确的 API 契约、验证结果、遵守配额，并保留时间戳和来源记录。