Free FX Data API
Selection and Integration Guide免费外汇数据 API
选择与接入指南

Free FX data API: what to look for first免费外汇数据 API：先检查哪些条件

Define the FX dataset, not just one exchange-rate endpoint先定义完整 FX 数据集，而不是只找一个汇率端点

Spot rates are only one layer of foreign-exchange data即期汇率只是外汇数据的一层

This page is for research and data teams comparing broad FX coverage. Depending on the project, “FX data” may include spot reference rates, historical closes, intraday bars, bid/ask quotes, central-bank fixings, forward points, outright forwards, swaps, implied yields, volatility measures, and macro series. List the required instruments and tenors before evaluating a provider; otherwise a good conversion API can be mistaken for a complete research dataset.

本页面向比较广义外汇覆盖的研究人员和数据团队。根据项目不同，“FX 数据”可能包括即期参考汇率、历史收盘、盘中 K 线、买卖报价、央行定盘、远期点、远期汇率、掉期、隐含收益率、波动率指标和宏观序列。评估供应商前应先列出所需工具与期限，否则很容易把一个好用的换算 API 误认为完整研究数据集。

Model instrument, tenor, fixing, and observation separately分别建模工具、期限、定盘与观测

A normalized record should identify the currency pair, instrument type, tenor or maturity, quote convention, source, fixing name where relevant, observation time, publication time, timezone, rate side, and unit. “EURUSD 1M” can mean a forward point, outright forward, volatility tenor, or aggregation interval. Do not overload one symbol field with all of these meanings.

标准化记录应明确币种对、工具类型、期限或到期日、报价惯例、来源、相关定盘名称、观测时间、发布时间、时区、报价方向和单位。“EURUSD 1M”可能指一个月远期点、远期汇率、波动率期限，也可能只是聚合周期，不能把这些含义都塞进一个 symbol 字段。

Worked forward-points example: tenor and units change the meaning远期点案例：期限与单位会改变报价含义

Assume a synchronized snapshot has EUR/USD spot at 1.08500 and a one-month mid forward adjustment of +18.5 pips. Under a four-decimal pip convention, the adjustment is 18.5 × 0.0001 = 0.00185, so the one-month outright forward is 1.08500 + 0.00185 = 1.08685. A provider may instead transmit the points as the integer 185 and document that each unit represents 0.00001. Adding the raw value 185 to spot is obviously wrong; treating 185 as 185 pips gives 1.10350 and is still wrong by a factor of ten.

Store instrument_type=forward_points, tenor, spot date, maturity or value date, day-count and holiday-calendar context, side, point scale, spot source, observation time, and whether the outright was supplied or derived. Bid and ask outrights must be built from compatible sides rather than one midpoint. A “1M” tenor is not always a fixed number of calendar days: broken dates, month-end rules, holidays, and settlement conventions can change the value date. Historical research should therefore join by the actual maturity and dataset vintage, not by a label alone.

远期点案例：期限与单位会改变报价含义

假设同一时点的 EUR/USD 即期中间价为 1.08500，一个月远期调整为 +18.5 个 pip。按照四位小数的 pip 规则，调整值为 18.5 × 0.0001 = 0.00185，因此一个月远期汇率为 1.08500 + 0.00185 = 1.08685。但供应商也可能把远期点传成整数 185，并在文档中规定每个单位代表 0.00001。把原始数值 185 直接加到即期价显然错误；把它误读成 185 个 pip，得到 1.10350，同样会产生十倍误差。

记录中应明确保存 instrument_type=forward_points、期限、即期交割日、到期或起息日、日计数与节假日日历背景、买卖方向、点值缩放、即期来源、观测时间，以及远期汇率是供应商直接返回还是本地推导。远期买价和卖价必须由兼容的同侧报价构造，不能只使用一个中间价。“1M”也不一定等于固定日数：非标准到期日、月末规则、节假日和结算惯例都可能改变实际起息日。因此，历史研究应按真实到期日和数据版本关联，不能只靠期限标签。

Coverage quality includes history, revisions, and missing regimes覆盖质量还包括历史深度、修订规则与缺失制度

Research datasets need stable backfills and documented revision behavior. Check the earliest date by instrument, holiday calendars, timezone changes, currency redenominations, discontinued benchmarks, synthetic crosses, and whether historical values can be revised after publication. A broad pair count is less valuable than a smaller universe with consistent lineage and no unexplained gaps.

研究数据集需要稳定回补和明确修订规则。应检查各类工具的最早日期、节假日日历、时区变化、货币重估、停用基准、合成交叉盘，以及历史数值发布后是否会修订。数量很多但来源不清的币种对，不如覆盖较小却口径一致、没有无解释缺口的数据集可靠。

Compare free FX datasets by research coverage and lineage按研究覆盖与数据来源链比较免费 FX 数据

Use the free layer to audit coverage, not to assume production economics用免费层审计覆盖，不要据此推断生产成本

A free plan is valuable for checking symbols, units, timestamps, gap behavior, and sample history. It rarely proves that a full multi-year, multi-instrument backfill is economical. Estimate rows, calls, export size, and refresh frequency for the real universe; then verify bulk-download rights, storage permissions, and redistribution terms before treating the API as a data-platform dependency.

免费套餐很适合检查代码、单位、时间戳、缺口行为和样例历史，但通常不能证明多年、多工具回补在经济上可行。应按真实标的范围估算行数、调用量、导出体积和刷新频率，再确认批量下载权、存储权限和再分发条款，之后才能把它作为数据平台依赖。

Choose the right FX data source and rate type选择正确的外汇数据源与汇率类型

API features are only useful when the underlying rate matches the job. Before comparing request limits or SDKs, identify whether the provider returns an official reference series, an aggregated midpoint, a delayed market snapshot, or venue-specific bid and ask prices. Two valid APIs can disagree because they observe different markets, fixing times, contributors, and calculation rules.

只有底层汇率口径符合任务，API 功能才有意义。比较请求额度或 SDK 前，应先确认供应商返回的是官方参考序列、聚合中间价、延迟市场快照，还是特定交易场所的买卖价。两个都有效的接口可能因为市场、定盘时间、报价来源和计算规则不同而出现差异。

Midpoint, bid, ask, open, high, low, and close中间价、买价、卖价与 OHLC

A single rate field is ambiguous unless the documentation defines it. A trading-oriented feed may provide bid, ask, midpoint, and OHLC values for an interval, while a conversion API may expose one reference value. Store the rate type with every observation so downstream code does not treat a midpoint as an executable price.

如果文档没有定义，单独一个 rate 字段存在歧义。交易类数据源可能提供买价、卖价、中间价以及某个周期的 OHLC，而换算接口可能只返回一个参考值。应为每条观测保存汇率类型，避免下游代码把中间价当成可成交价格。

Base currency restrictions and cross-rate calculations基准币限制与交叉汇率计算

Some free plans allow only one base currency. Converting through that base can expand coverage, but it introduces two observations, extra rounding, and a dependency on consistent timestamps. Record whether a pair was supplied directly or derived, the calculation path, and the precision used.

部分免费套餐只允许一个基准币。通过该基准币换算可以扩大覆盖，但会引入两条观测、额外舍入，并依赖时间戳一致性。应记录币种对是直接返回还是推导得到、计算路径以及采用的精度。

Test a free FX data API JSON response测试免费外汇数据 API 的 JSON 响应

Provider-neutral latest-rates request与供应商无关的最新汇率请求

GET /latest?base=USD&symbols=EUR,JPY
Accept: application/json

{
  "base": "USD",
  "timestamp": "SOURCE_TIMESTAMP",
  "rates": {
    "EUR": "RATE_VALUE",
    "JPY": "RATE_VALUE"
  }
}

The endpoint and values are placeholders, not claims about a provider or current market rates. Replace them with the selected API documentation, then verify direction, timestamp, numeric type, missing symbols, and error behavior.

端点与数值均为占位示例，不代表任何供应商或当前市场汇率。请按选定接口文档替换，并校验方向、时间戳、数值类型、缺失币种和错误行为。

Fields a production response should make unambiguous生产响应中应明确的字段

Integration checklist for a free currency API免费汇率接口接入清单

1. Validate base and quote codes before sending the request.
2. Set an explicit timeout and handle 4xx, 5xx, and rate-limit responses separately.
3. Store the source timestamp, retrieval time, provider, and raw response checksum.
4. Use decimal arithmetic and make cache or database writes idempotent.
5. Alert on stale timestamps, schema changes, missing pairs, and quota exhaustion.

1. 发送请求前校验基准币与目标币代码。
2. 设置明确超时，分别处理 4xx、5xx 与限频响应。
3. 保存来源时间、抓取时间、供应商与原始响应校验值。
4. 使用十进制运算，并让缓存或数据库写入具备幂等性。
5. 对陈旧时间戳、结构变化、缺失币种对与额度耗尽告警。

Validate FX data before production use外汇数据用于生产前的校验

Reference rates are not executable prices参考汇率不是可成交价格

A midpoint or official reference rate does not include a bank spread, card fee, slippage, or a venue-specific bid and ask. Label the data accurately and do not present it as an execution guarantee.

中间价或官方参考价不包含银行点差、银行卡费用、滑点或特定交易场所的买卖价。应准确标注数据口径，不能将其描述为成交保证。

Freshness, continuity, and outlier checks时效、连续性与异常值检查

Reject non-positive or malformed rates, detect stale source timestamps, compare reciprocal pairs within a defined tolerance, and classify missing observations. A second source can trigger review, but should not silently manufacture a replacement value.

拒绝非正数或异常结构，检测陈旧来源时间，在设定容差内比较倒数币种对，并分类缺失观测。第二来源可触发复核，但不应静默创造替代值。

Quota-aware caching and refresh design感知额度的缓存与刷新设计

Match cache duration to the provider’s actual update schedule. Calling a daily reference endpoint every minute wastes quota without improving freshness. Keep a short-lived stale-on-error window for non-critical displays, but expose the observation timestamp so users and downstream services can distinguish cached data from a fresh response.

缓存时间应与供应商的实际更新计划匹配。每分钟调用一次日频参考接口只会浪费额度，不会提升时效。非关键展示可在接口出错时短暂沿用旧数据，但必须展示观测时间，让用户和下游服务能够区分缓存数据与最新响应。

Retries, circuit breakers, and provider fallback重试、熔断与供应商回退

Retry only transient failures with exponential backoff and jitter. Do not retry invalid currencies, authentication failures, or exhausted monthly quotas as if they were network timeouts. A fallback provider must be normalized to the same pair direction, timestamp rules, and rate type before its values can replace the primary source.

只对暂时性故障使用带随机抖动的指数退避。无效币种、认证失败或月度额度耗尽不能当作网络超时反复重试。备用供应商的数据必须先统一币种对方向、时间戳规则和汇率类型，才能替代主数据源。

Licensing, attribution, and redistribution许可、署名与再分发

Free access to an endpoint does not automatically grant the right to cache indefinitely, publish a bulk download, resell the data, or remove attribution. Keep the terms version and review date with the integration record, and re-check them before changing how the data is displayed or distributed.

免费访问端点并不自动授予无限期缓存、公开批量下载、转售数据或移除署名的权利。应在接入记录中保存条款版本和复核日期，并在改变数据展示或分发方式前重新检查。

How QVeris helps you find a free FX data APIQVeris 如何帮助查找免费外汇数据 API

QVeris helps developers and agents discover and inspect callable capabilities. Open the QVeris tool details to inspect relevant capabilities, inspect their inputs and outputs, then follow the QVeris documentation to connect them. QVeris can reduce discovery and schema-inspection work; it does not change a provider’s rates, license, quota, or service level.

QVeris 帮助开发者与 Agent 发现并检查可调用能力。可打开 QVeris 工具详情，查看相关能力，检查输入输出，再按 QVeris 文档完成接入。QVeris 能减少发现与结构检查工作，但不会改变供应商的汇率、许可、额度或服务水平。

Free FX API questions免费外汇 API 常见问题