Choose a Free Financial API
Data, Sources & Examples选择合适的免费金融 API
数据、来源与示例

Quick Summary快速摘要

Choose a free financial API by use case按使用场景选择免费金融 API

The best API is the one whose data contract matches the decision your product must make. A dashboard, a valuation notebook, a backtest, and an AI research agent can all request “financial data,” yet they require different timestamps, identifiers, update schedules, and usage rights. Define the output first, then work backward to the source.

最合适的 API，应当让数据契约与产品要完成的决策相匹配。看板、估值 Notebook、回测系统和 AI 研究 Agent 都会请求“金融数据”，但它们需要的时间戳、标识符、更新频率和使用权并不相同。应先定义输出，再反推所需数据源。

A practical shortlist should contain no more than two or three providers for the first test. If the job is daily company research, prioritize filings, fundamentals, corporate actions, and batch endpoints. If the job is an intraday alert, prioritize feed coverage, latency, connection behavior, and the license for displaying results. A provider can be excellent for one workflow and unsuitable for the other.

第一轮测试最好只保留两到三个候选供应商。日频公司研究应优先检查申报文件、基本面、公司行动和批量接口；日内提醒则应优先检查行情源覆盖、延迟、连接行为以及展示结果所需的许可证。同一供应商可能非常适合一种流程，却完全不适合另一种流程。

Which financial API endpoints do you need?你需要哪些金融 API 接口？

Start with the decision or product you are building. A portfolio app needs current prices; a valuation model needs fundamentals; macro research needs rates, inflation, employment, and GDP; an agent workflow needs stable schemas and source metadata.

先明确你的决策或产品目标。投资组合应用需要当前价格，估值模型需要基本面，宏观研究需要利率、通胀、就业与 GDP，Agent 工作流则需要稳定字段与来源元数据。

A free financial API is rarely one endpoint. Market routes may return quotes or interval bars, company routes follow reporting cycles, and economic series follow official release calendars. Define the asset class, geography, frequency, history depth, acceptable delay, and response format before comparing providers.

免费的金融 API 通常不止一个接口。市场接口返回行情或周期数据，公司接口跟随财报周期更新，经济指标则依据官方发布日历更新。比较供应商之前，应先明确资产类别、地区、频率、历史深度、可接受延迟和响应格式。

For research, price fields alone are rarely enough. You may also need trading calendars, symbol changes, delistings, currency, exchange identifiers, and corporate actions. These reference fields prevent survivorship bias and make it possible to join prices with fundamentals correctly. A smaller dataset with explicit lineage and consistent identifiers is often more useful than a larger feed whose origin and adjustment rules are unclear.

对于研究任务，只有价格字段通常不够。你还可能需要交易日历、代码变更、退市记录、币种、交易所标识和公司行动。这些参考字段能减少幸存者偏差，也能让价格与基本面正确关联。来源清楚、标识一致的小型数据集，往往比来源不明、复权规则模糊的大型数据源更有价值。

How free financial APIs compare如何比较免费的金融 API

No single source is best for every workflow. Exchanges and regulators are closest to primary records, but their files may require more cleaning. Market-data APIs make common fields easier to query, although free plans usually trade convenience for lower request limits, delayed prices, shorter history, or narrower exchange coverage. Broker feeds can be useful for personal execution workflows, while downloadable files are efficient for reproducible analysis.

没有一种来源适合所有工作流。交易所和监管机构最接近一手记录，但文件往往需要更多清洗；市场数据 API 查询方便，不过免费套餐通常以较低调用限额、延迟行情、较短历史或较窄市场覆盖来换取便利。券商数据适合个人交易流程，而可下载文件更适合可重复的研究分析。

Compare sources with the same small test set rather than relying on feature lists. Request a liquid stock, a thinly traded security, a symbol that changed, and a date containing a split or dividend. Compare timestamps, missing records, precision, adjustment behavior, and error messages. This test quickly reveals whether two providers use different sessions or definitions even when their column names look identical.

不要只看功能清单，最好用同一组小样本横向测试：选择一只高流动性股票、一只成交稀疏的证券、一个曾更换代码的标的，以及包含拆股或分红的日期。对比时间戳、缺失记录、数值精度、复权方式和错误信息。即使字段名称相同，这种测试也能快速发现不同供应商对交易时段和指标定义的差异。

How to evaluate a free financial API before integration接入前如何评估免费金融 API

Feature lists are easy to copy and hard to compare. A better evaluation uses the same representative requests, expected outputs, and failure tests for every candidate. Record each result in a scorecard so “free,” “real-time,” and “global” are translated into testable definitions rather than marketing labels.

功能清单容易复制，却很难直接比较。更可靠的评估方法，是对每个候选 API 使用相同的代表性请求、预期输出和失败测试，并把结果记录在评分表中。这样，“免费”“实时”和“全球覆盖”就会变成可验证的定义，而不是营销标签。

1. Coverage and identifiers1. 覆盖范围与标识符

List the exact exchanges, countries, asset classes, security types, and symbol states you need. Test active and delisted instruments, primary and secondary listings, ETFs, preferred shares, and symbols that changed. Prefer permanent identifiers or explicit mapping tables over ticker-only joins, because tickers can be reused or differ by venue.

列出真正需要的交易所、国家、资产类别、证券类型和代码状态。测试正常上市与退市标的、主次上市、ETF、优先股以及曾经变更代码的证券。连接数据时，应优先使用永久标识符或明确的映射表，而不是只靠股票代码，因为代码可能被重复使用，也可能因交易场所不同而变化。

2. Freshness, timestamps, and market sessions2. 时效、时间戳与交易时段

Ask what each timestamp represents: event time, exchange time, bar close, provider ingestion, or delivery time. Verify regular, pre-market, and after-hours sessions separately. “Real-time” describes latency, not necessarily complete market coverage; a live single-venue feed and a consolidated feed can produce different prices and volumes.

要明确每个时间戳表示事件发生时间、交易所时间、K 线结束时间、供应商入库时间，还是发送时间。盘前、正常交易和盘后时段应分别验证。“实时”描述的是延迟，并不必然代表完整市场覆盖；实时单一场所行情与综合行情可能返回不同的价格和成交量。

3. Point-in-time quality and revisions3. 时点数据质量与修订

For fundamentals and economics, determine whether the API preserves the value known on a past date or only serves the latest revised value. Record filing dates, accepted timestamps, fiscal periods, units, and revision or restatement status. Point-in-time data is essential when a model claims to reproduce a historical decision.

对于基本面和宏观数据，要判断 API 是否保留历史某一时点已知的数值，还是只提供最新修订值。需要记录申报日期、接收时间、财务期间、单位以及修订或重述状态。只要模型声称能够重现历史决策，时点数据就是必要条件。

4. Reliability and failure behavior4. 可靠性与失败行为

Test invalid symbols, missing dates, quota exhaustion, timeouts, and partial responses. The client should distinguish “no observation exists” from “the provider failed.” Capture HTTP status, retry headers, provider error messages, and the age of cached data. Silent fallback to stale or zero values is more dangerous than an explicit error.

应测试无效代码、缺失日期、额度耗尽、超时和部分返回。客户端必须区分“确实没有观测值”和“供应商请求失败”。记录 HTTP 状态、重试响应头、供应商错误信息和缓存数据的新旧程度。在没有提示的情况下改用过期数据或零值，往往比明确报错更危险。

5. Licensing and permitted outputs5. 许可与输出权限

Separate personal analysis, internal business use, public display, non-display use, derived data, model training, and redistribution. A working API key proves technical access, not permission to publish or resell the result. Save the applicable plan and license URL alongside the integration decision, then review it when the product or audience changes.

个人分析、企业内部使用、公开展示、非展示用途、衍生数据、模型训练和再分发应分别核对。API Key 能正常调用，只能证明技术访问成功，并不能证明可以发布或转售结果。应把适用套餐和许可链接与接入决策一起保存，并在产品或用户群发生变化时重新检查。

A practical financial API integration workflow实用的金融 API 接入流程

A reliable workflow separates discovery from production use. First inspect the provider documentation and a sample response. Then make a small request, save the raw payload, and transform it into your own stable schema. Add validation before the data enters a chart, model, spreadsheet, or agent. This makes it easier to replace a source later without rewriting the rest of the product.

可靠的工作流会把“发现数据源”和“正式使用”分开。先查看供应商文档与示例响应，再发起小规模请求，保存原始返回，并转换为自己的稳定字段结构。数据进入图表、模型、表格或 Agent 之前应先经过验证。这样未来更换数据源时，无需重写整个产品。

For API access, cache responses that do not need second-by-second freshness, apply exponential backoff to temporary failures, and log the request time, parameters, status code, and provider timestamp. For CSV, preserve the original file separately from the cleaned table and record its download date, delimiter, encoding, and time zone. Both approaches need a repeatable update policy so old and new observations are not silently mixed.

使用 API 时，应缓存不要求秒级更新的结果，对临时失败采用指数退避，并记录请求时间、参数、状态码和供应商时间戳。使用 CSV 时，应把原始文件与清洗后的表格分开保存，同时记录下载日期、分隔符、编码和时区。两种方式都需要可重复的更新策略，避免新旧数据在无提示的情况下混合。

Free financial API Python example with validation带验证步骤的免费金融 API Python 示例

A useful first script should do more than print a value. The example below requests company facts from the SEC EDGAR API, stores the source and retrieval time, checks the response type, and exposes the filing metadata that helps an analyst interpret a number. SEC documentation states that data.sec.gov provides JSON APIs without an API key; automated access must still follow the SEC's access policy and identify the requester.

有用的第一段脚本不应只打印一个数值。下面的示例从 SEC EDGAR API 请求公司事实数据，保留来源和获取时间，检查返回类型，并展示帮助分析师解释数值的申报元数据。SEC 文档说明，data.sec.gov 提供无需 API Key 的 JSON API；自动访问仍须遵守 SEC 访问政策，并正确标识请求者。

import os
from datetime import datetime, timezone

import requests

cik = "0000320193"  # Apple Inc.; CIK must contain 10 digits
url = f"https://data.sec.gov/api/xbrl/companyfacts/CIK{cik}.json"
headers = {
    "User-Agent": os.environ["SEC_USER_AGENT"],
    "Accept-Encoding": "gzip, deflate",
}

response = requests.get(url, headers=headers, timeout=20)
response.raise_for_status()
payload = response.json()

if "facts" not in payload or "entityName" not in payload:
    raise ValueError("Unexpected SEC response schema")

revenue = payload["facts"]["us-gaap"]["RevenueFromContractWithCustomerExcludingAssessedTax"]
recent_usd_facts = revenue["units"]["USD"][-5:]

result = {
    "source": "SEC EDGAR Company Facts",
    "retrieved_at": datetime.now(timezone.utc).isoformat(),
    "entity": payload["entityName"],
    "taxonomy_label": revenue["label"],
    "recent_facts": recent_usd_facts,
}

print(result)

Set SEC_USER_AGENT to the application name and a monitored contact address before running the script. In production, do not assume the last array item is the “correct” value. Filter by form, fiscal period, filing date, unit, frame, and accepted timestamp, then decide how amendments and duplicate facts should be handled. For large-scale extraction, use the SEC bulk archives instead of repeating one-company calls.

运行前，将 SEC_USER_AGENT 设置为应用名称和可联系的邮箱。生产环境中不能直接假设数组最后一项就是“正确值”，而应按表单、财务期间、申报日期、单位、frame 和接收时间过滤，并明确如何处理修订文件和重复事实。大规模提取时，应使用 SEC 批量归档，而不是持续重复调用单家公司接口。

Model the real request budget and hidden cost计算真实请求预算与隐藏成本

A headline quota such as requests per minute or requests per day is not enough to predict whether a free tier will work. Calculate the full path required to produce one answer:

仅看每分钟或每天请求次数，无法判断免费套餐是否够用。应计算生成一次完整答案所需的全部调用路径：

Suppose a watchlist contains 25 companies and each refresh needs a quote, company profile, income statement, ratios, and recent news. Four refreshes per day already require 500 calls before retries. A 10% retry allowance raises the budget to 550, while exchange calendars, FX conversion, and symbol mapping add shared calls. Batch endpoints, caching, and separating slow-changing fundamentals from fast-changing prices can reduce the total dramatically.

假设观察列表有 25 家公司，每次刷新需要行情、公司资料、利润表、财务比率和近期新闻五个接口，每天刷新四次，在重试前就需要 500 次调用。预留 10% 重试空间后变成 550 次，交易日历、汇率换算和代码映射还会增加公共调用。批量接口、缓存，以及把低频变化的基本面与高频变化的价格分开更新，都能显著降低请求量。

Costs that do not appear in the API priceAPI 标价之外的成本

• Normalization: mapping provider-specific fields, identifiers, currencies, time zones, and null conventions into a stable internal schema.
• Quality control: detecting gaps, duplicates, revisions, stale observations, incorrect adjustment behavior, and unexpected outliers.
• Operations: caching, retry queues, monitoring, provider-status checks, key rotation, and incident investigation.
• Licensing: obtaining the rights needed for public display, professional users, derived outputs, or redistribution.

• 标准化：把供应商特有字段、标识符、币种、时区和空值规则映射为稳定的内部结构。
• 质量控制：发现缺口、重复、修订、过期观测、错误复权和异常值。
• 运行维护：缓存、重试队列、监控、供应商状态检查、密钥轮换和故障调查。
• 许可：取得公开展示、专业用户、衍生输出或再分发所需的权利。

Create an upgrade trigger before launch. Typical triggers include sustained usage above a safe share of the quota, a need for lower latency or consolidated coverage, insufficient history, missing corporate actions, or a customer-facing license requirement. The decision should be tied to product behavior and risk—not an arbitrary date.

上线前应提前定义升级触发条件，例如正常使用量持续接近安全额度上限、需要更低延迟或综合行情、历史深度不足、缺少公司行动，或需要面向客户的许可。升级决策应由产品行为和风险触发，而不是设定一个随意日期。

Run a 15-minute free-tier fit test用 15 分钟判断免费套餐是否真的适合

A short, repeatable smoke test is more useful than browsing another feature list. It will not prove production reliability, but it can expose the most common reasons a free financial API fails a real workload: incomplete coverage, ambiguous timestamps, unstable identifiers, silent nulls, unusable quotas, and rights that do not match the product.

与其继续浏览功能清单，不如做一次可重复的小型冒烟测试。15 分钟当然不能证明接口已经达到生产可靠性，但足以暴露最常见的问题：覆盖不完整、时间戳含义不清、证券标识不稳定、空值没有说明、额度无法支撑实际负载，或使用许可与产品场景不匹配。

Free financial API checklist免费金融 API 检查清单

Data quality is not only about whether a number looks plausible. Confirm that every observation refers to the intended venue, session, currency, and time zone. Check whether timestamps represent trade time, bar close, or provider delivery time. When building indicators or backtests, document how missing sessions, zero volume, duplicate rows, outliers, splits, dividends, and restated fundamentals are handled.

数据质量不只是判断一个数字“看起来是否合理”。每条记录都应对应正确的交易场所、交易时段、币种和时区，还要确认时间戳表示成交时间、K 线结束时间，还是供应商发送时间。构建指标或回测时，应记录如何处理缺失交易日、零成交量、重复行、异常值、拆股、分红以及重述后的财务数据。

• Confirm the market, exchange, asset class, interval, and historical depth.
• Record whether prices are raw or adjusted and how corporate actions are handled.
• Preserve source, timestamp, currency, time zone, and request parameters with every result.

• 确认市场、交易所、资产类别、周期和历史深度。
• 记录价格是否复权，以及拆股和分红等公司行动的处理方式。
• 为每条结果保留来源、时间戳、币种、时区和请求参数。

Finally, read the license as carefully as the schema. Free access may permit personal analysis but restrict public display, redistribution, automated extraction, model training, or commercial products. Keep a link or copy of the applicable terms with the project and review them when the provider changes plans. If the data supports trading, compliance, or customer-facing decisions, validate critical values against an authoritative source.

最后，许可证应当和字段说明一样认真阅读。免费访问可能允许个人分析，却限制公开展示、再分发、自动抓取、模型训练或商业产品。建议在项目中保存适用条款的链接或副本，并在供应商调整套餐时重新检查。如果数据用于交易、合规或面向客户的决策，关键数值还应与权威来源交叉核验。

FAQ常见问题

External references外部参考链接