This server provides AI agents with two stdio MCP tools for accessing daily gold and macro readings from xaudaily.com.
get_gold_readings: returns full structured JSON readings (schema xaudaily.readings/v1) covering COMEX gold, SHFE Au99.99, US CPI/core PCE/PPI, nonfarm payrolls, DXY, Treasury 10Y/30Y yields, VIX, SPDR gold ETF holdings, fed funds rate, Polymarket FOMC odds, central-bank gold buying, Brent/WTI, US debt, and a gold driver score with event estimates.
Supports optional fields=[...] to fetch only selected top-level keys or dotted subfields (e.g., ["gold","cpi","extra.polymarket"]) to save context.
Each field includes its own source and asOf/date, and is flagged stale: true when an upstream source fails, so citations are verifiable.
get_gold_daily_brief: returns today's key numbers as a short Markdown brief (a few KB), including gold, SHFE gold, CPI, core PCE, NFP, PPI, DXY, Treasury 10Y, Brent, fed funds rate, next-FOMC pricing, central-bank gold buying, US debt, and top drivers.
Updated twice daily (06:30 / 22:40 JST) with a ~30-minute gold tick; it is daily readings, not a real-time feed.
Attribution is required when citing: “Source: Gold Data Reading · XAU Daily”.
The server uses only the Python standard library, speaks JSON-RPC over stdio, and returns structured errors on timeouts/HTTP failures without crashing.
An Agent Skill plus a zero-dependency MCP server that give AI agents daily gold and macro readings from xaudaily.com — numbers an agent can actually cite.
Why not just let the agent search the web? Every field carries its own source and asOf/date, so a
citation can be checked rather than trusted. When an upstream source fails, the site keeps the last good value
and flags the field stale: true instead of silently serving an old number. And the figures are the same ones
the site renders — same snapshot, same code path — so the JSON and the page can never disagree.
Two tools, over stdio:
Tool | Returns |
| Full JSON readings (schema |
| Today's headline numbers as Markdown (a few KB) |
What's inside: COMEX gold and Shanghai Au99.99 (with a ~30-minute live tick), US CPI / core PCE / nonfarm payrolls / PPI, DXY, Treasury 10Y and 30Y yields, VIX, SPDR gold ETF holdings, the fed funds rate and Polymarket odds for the next FOMC decision, IMF central-bank gold buying, Brent/WTI crude, US debt, plus a rule-based gold driver score and event estimates — one request, about 60 KB, instead of stitching together a dozen sources.
Updated twice a day (06:30 and 22:40 JST) with a ~30-minute gold tick. These are daily readings, not a real-time feed — please don't describe them as real-time or guaranteed accurate. The data is not investment advice.
Files: SKILL.md (the skill), mcp_server.py (the server — standard library only), Dockerfile (for
container-based catalogs), glama.json, examples/ (copy-paste curl and Python), WIDGET.md (an embeddable
gold badge).
As an MCP server. One Python file, nothing to pip install:
{
"mcpServers": {
"xaudaily": {
"command": "python3",
"args": ["/absolute/path/to/mcp_server.py"],
"env": {
"XAUDaily_READINGS_URL": "https://xaudaily.com/readings.en.json?src=skill-mcp",
"XAUDaily_BRIEF_URL": "https://xaudaily.com/brief.en.md?src=skill-mcp"
}
}
}
}With Docker (the same image MCP catalogs build for introspection):
docker build -t xaudaily-mcp .
docker run -i --rm xaudaily-mcp # then speak JSON-RPC on stdinAs an Agent Skill: copy this directory (or just SKILL.md) into your runtime's skills directory.
English payload: https://xaudaily.com/readings.en.json?src=skill-github
Chinese payload: https://xaudaily.com/readings.json?src=skill-github — same snapshot, same numbers, same
field paths; only the string values (units, sources, caveats) are Chinese
Briefs: /brief.en.md and /brief.md
Data is CC BY 4.0. Attribution with a link is required — e.g. Source: Gold Data Reading · XAU Daily. Please don't repackage the dataset as your own. For commercial use or redistribution, contact xaudaily@163.com.
中文说明见下方 ↓
Related MCP server: Wicked API - Agent First API Services
给 AI Agent 用的黄金宏观数据接入包：一份 Agent Skill 说明书 + 一个零依赖的 MCP server。
数据源：xaudaily.com（品牌「黄金读数 / Gold Data Reading · XAU Daily」），纯静态黄金宏观数据仪表盘
端点（英文）：https://xaudaily.com/readings.en.json?src=skill-github（JSON，schema xaudaily.readings/v1）
端点（中文）：https://xaudaily.com/readings.json?src=skill-github —— 同一份快照、同一批数字，
字段路径完全一致，只是字符串值（单位/来源/口径说明）为中文；英文简报 /brief.en.md，中文简报 /brief.md
许可：CC-BY-4.0（署名 + 附链接）
skill-repo/
├── SKILL.md # Agent Skill：字段说明、信封约定、常见坑（中文为主，关键段落英文）
├── mcp_server.py # MCP server：纯 Python 标准库，stdio 传输，两个工具
├── Dockerfile # 极简镜像，给 Glama 这类按容器做 introspection 的目录用
├── glama.json # 声明维护者（Glama 的收录元数据）
├── README.md # 本文件（英文在前，中文见本节）
├── LICENSE # CC-BY-4.0 完整法律文本
├── WIDGET.md # 可嵌入金价徽标的嵌入说明（给你的网站/README 用）
└── examples/
├── curl.md # 可直接复制运行的 curl 示例
└── python.md # 只用标准库的 Python 示例xaudaily.com 每天 06:30 与 22:40（JST）自动抓取、校验并渲染一批黄金宏观读数，同时把它输出成一份给机器读的 JSON。这个仓库把那份额外的机器可读产出包装成 Agent 能直接消费的形态：
SKILL.md —— 一份 Agent Skills 约定的说明书：端点表、完整字段结构（异构结构逐个列出）、取数示例、数据口径、更新节奏、署名要求、踩过的坑。
mcp_server.py —— 支持 MCP 的客户端（Claude Desktop 等）挂上就能用的 server，两个工具：get_gold_readings（全量 JSON，可按字段裁剪）与 get_gold_daily_brief（当日 Markdown 简报）。只用 Python 标准库，不装任何第三方包。
每个字段自带出处与截止日。 每个字段里有 source（数据来源）和 asOf/date（数据截止日），meta 里还有口径说明、免责声明与署名要求。Agent 引用时能给出可核查的出处，而不是一个来路不明的数字。
降级是显式的。 某个上游源抓取失败时，站点沿用上一次成功值并把该字段标成 stale: true —— 消费方能看出「这不是最新一期」，而不是被悄悄喂了旧数字。
字段稳定、结构有版本。 顶层 schema 恒为 xaudaily.readings/v1，改结构时会换值；extra 区可以加新字段而不动顶层键，所以你的代码不容易因为上游加字段而挂掉。
单次抓取就够。 金价、沪金、CPI、PCE、非农、PPI、DXY、美债 10Y/30Y、VIX、SPDR 持仓、联邦基金利率、FOMC 决议概率、央行购金、原油、美国债务、驱动因子评分与事件预判，全在一次请求里（约 60 KB），不用拼十几个来源。
零依赖 = 不会腐烂。 mcp_server.py 是单个 stdlib-only 的 Python 文件，没有 requirements，几年后照样能跑。
把本目录（或其中的 SKILL.md）放进你的 Agent 运行时的 skills 目录即可 —— Agent Skills 约定是「一个目录一份技能，入口是 SKILL.md」：
git clone <本仓库地址> gold-market-readings
cp -r gold-market-readings ~/.your-agent/skills/gold-market-readings/
# 或者直接把 SKILL.md 放到你现有的 skills 目录下Agent 会读它的 frontmatter（name / description）判断何时该用：涉及金价、通胀（CPI/PCE）、非农、FOMC 决议概率、央行购金、美债收益率、DXY、黄金驱动因子时触发。
mcp_server.py 走 stdio 传输，不需要端口、不需要守护进程。把它加进客户端的 MCP 配置：
{
"mcpServers": {
"xaudaily": {
"command": "python3",
"args": ["/absolute/path/to/mcp_server.py"],
"env": {
"XAUDaily_READINGS_URL": "https://xaudaily.com/readings.en.json?src=skill-mcp",
"XAUDaily_BRIEF_URL": "https://xaudaily.com/brief.en.md?src=skill-mcp"
}
}
}
}Windows 上把 command 写成解释器的完整路径最稳妥。手动试一下：
python3 mcp_server.py # 它会等待 stdin 上的 JSON-RPC 报文环境变量 | 默认值 | 用途 |
|
| 覆盖读数端点（默认值＝英文端点，已带渠道参数） |
|
| 覆盖简报端点（默认值＝英文端点） |
| 空 | 指向本地一份 |
|
| 单次 HTTP 超时秒数 |
工具的行为约定（也是 SKILL.md 里写给 Agent 的部分）：
未知方法返回 JSON-RPC error（-32601），不会崩；notifications/* 一律不响应。
网络超时、HTTP 非 200、JSON 解析失败，都转成结构化的 tool error 结果（isError: true + kind/detail/hint），进程照常活着 —— 一次网络抖动不该让 Agent 的整个会话失去这个工具。
stdout 上只有 JSON-RPC 报文；日志一律走 stderr。
仓库根目录的 Dockerfile 就是给容器化场景用的（Glama 这类目录按它构建镜像、启动进程、读 tools/list 做 introspection）：
docker build -t xaudaily-mcp .
docker run -i --rm xaudaily-mcp # 然后往 stdin 写 JSON-RPC# 全量读数
curl -sS 'https://xaudaily.com/readings.json?src=skill-github' -o readings.json
# 只要今天的关键数字（Markdown，几 KB）
curl -sS 'https://xaudaily.com/brief.md?src=skill-github'import json, urllib.request
with urllib.request.urlopen("https://xaudaily.com/readings.json?src=skill-github", timeout=20) as r:
d = json.load(r)
print(d["generated_at"], "| 数据截止", d["data_asof"])
print("沪金", d["readings"]["au"]["last"], d["readings"]["au"]["unit"])MCP 侧一次 tools/call（收到的是两个 content 块：署名提醒 + 数据本体）：
{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"get_gold_readings","arguments":{"fields":["gold","extra.polymarket"]}}}更多可直接复制运行的例子见 examples/curl.md 与 examples/python.md。
?src=所有示例与文档里的 URL 都带 ?src=<渠道>，例如 ?src=skill-github、?src=skill-mcp。这不是跟踪参数，而是这个数据源能被持续维护的前提：
带 ?src= 的读数请求返回 Cache-Control: no-store，绕开 CDN 边缘缓存，每一次抓取都会真实落到源站访问日志；
不带参数时，响应走 Cache-Control: public, max-age=300，多次抓取会被边缘缓存合并成一次 —— 站长看到的就只有「3 次访问」，而实际可能是「1000 次消费」。
去掉参数请求不会失败，但渠道归因就此消失，也就没人说得清这份数据究竟有没有被用。所以：请不要删掉它；想区分你自己的渠道，把值改成你的名字（如 ?src=my-agent-prod）完全可以，只用 [a-z0-9-] 这类字符即可。
每天 两次全量更新：06:30 / 22:40（JST，东京时间；北京时间 05:30 / 21:40）。所有字段、存档页、brief.md 在这一刻刷新。
金价与沪金另外每约 30 分钟刷新一次实时 tick（只影响金价类字段）。
其余宏观指标是月频/季频数据（CPI、非农、GDP 等），本来就不每天变。
这不是行情终端。 请不要把它描述为「实时行情」或「保证准确」：数据由程序从公开来源抓取，可能滞后、被修订或缺失（缺失时看 stale 字段）。准确说法是「每日两次更新的宏观读数 + 约 30 分钟一次的金价 tick」。
许可：CC-BY-4.0（Creative Commons Attribution 4.0 International，SPDX：CC-BY-4.0，全文 https://creativecommons.org/licenses/by/4.0/）。
必须署名并附链接。 推荐写法：数据来源：黄金读数 xaudaily.com（英文：Source: Gold Data Reading · XAU Daily）。
不得把本站数据整体或实质部分再包装成「自有数据源」「自研 API」「独家数据」发布或转售。
商用、批量再分发、镜像、长期落库或作为训练语料：请先联系 xaudaily@163.com。
免责声明：数据由程序从公开来源抓取与呈现，不构成投资建议，不提供买卖指导；概率类字段是市场定价或规则模型估计，不是预测保证。
如果这份数据对你的读者有用，可以顺手嵌一枚每日自动更新的金价徽标：
<a href="https://xaudaily.com/?utm_source=widget&utm_medium=embed&utm_campaign=gold-badge">
<img src="https://xaudaily.com/widget/gold-badge.svg" width="320" height="96"
alt="今日金价 · 黄金读数 xaudaily.com">
</a>图由本站托管，每约 30 分钟跟着金价 tick 重生成，你不需要自己更新；
可点击的链接必须由外层 <a> 包裹 <img> 实现（<img> 加载的 SVG，其内部链接点不动）；请不要把图下载自托管——那样会永远停在旧价，也丢失归因；
深色站点用 /widget/gold-badge-dark.svg；配色是中国习惯（红涨绿跌），面向英文读者请在嵌入处补一句 "red = up"；
完整说明与三种嵌入写法见 WIDGET.md；在线预览：https://xaudaily.com/widget/
问题、字段需求、渠道归因（想让你自己的 agent 渠道单独出现在统计里）、商用授权：xaudaily@163.com
站点：https://xaudaily.com/ · Agent 自述文件：https://xaudaily.com/llms.txt?src=skill-github · 历史存档：https://xaudaily.com/d/?src=skill-github
[EN] Today's plain-text brief from xaudaily.com (Markdown, a few KB): today's gold, SHFE gold, CPI, core PCE, nonfarm payrolls, PPI, DXY, Treasury 10Y, Brent crude, the fed funds rate, next-FOMC market pricing, central-bank gold buying, US debt and the top gold drivers. Use it when you want today's headline numbers without parsing the full JSON. 取 xaudaily.com 当日纯文本简报（brief.md，Markdown）：当天金价、沪金、CPI、核心 PCE、非农、PPI、DXY、美债 10Y、布伦特原油、联邦基金利率、下次 FOMC 市场定价、央行购金、债务与驱动因子要点，一天一份，几 KB。需要快速拿到「今天的关键数字」而不想解析完整 JSON 时用它。内容每天 06:30 与 22:40（JST）刷新，因此同一天内多次调用可能拿到同一份内容。【引用要求（强制）】引用时注明来源并附链接，例如「数据来源：黄金读数 xaudaily.com」；不得再分发为自有数据源；商用请联系 xaudaily@163.com。Attribution is required: cite "Source: Gold Data Reading · XAU Daily".
| Name | Required | Description | Default |
|---|---|---|---|
No parameters | |||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations provided, the description carries the full behavioral burden and handles it well: it discloses format (Markdown, a few KB), source, refresh schedule, duplicate-content possibility, and mandatory attribution requirements. It does not mention failure modes or rate limits, but for a simple no-parameter read tool this is adequate.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is front-loaded with the format and content list, followed by usage timing and attribution rules. The bilingual repetition adds length, but both language blocks serve distinct audiences and the attribution clause is operationally necessary, so the extra text is justified.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
For a zero-parameter, no-output-schema tool, this description is complete: it specifies exactly what the brief contains, its size and format, when it refreshes, how to use it, and how to attribute the source. An agent can call and cite this tool correctly without needing additional context.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The tool has zero parameters, so the schema carries no semantic burden and the description correctly requires none. The 0-parameter baseline of 4 applies because nothing about parameters needs explanation.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description states a specific action and resource: retrieving the daily Markdown brief from xaudaily.com. It enumerates the exact content areas (gold, SHFE, CPI, PCE, nonfarm, etc.) and contrasts itself with parsing the full JSON, distinguishing it from sibling get_gold_readings.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
It explicitly says when to use this tool: when you want today's headline numbers without parsing the full JSON. Refresh times are given, and it notes repeated calls within the same day may return identical content. It does not explicitly name get_gold_readings as the alternative, but the full-JSON contrast conveys the decision boundary.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
[EN] Full machine-readable gold and macro readings from xaudaily.com (schema xaudaily.readings/v1): COMEX gold daily OHLC, SHFE Au99.99, US CPI / core PCE / PPI, nonfarm payrolls, DXY, Treasury 10Y/30Y yields, VIX, SPDR gold ETF holdings, the fed funds rate, Polymarket odds for the next FOMC decision, central-bank gold buying and reserves, Brent/WTI, US debt, and a rule-based gold driver score with event expectations. Every field carries its own source and asOf/date and is flagged stale=true when a source fails. Updated twice a day (06:30 / 22:40 JST) with a ~30-minute gold tick - daily readings, not a real-time feed; do not describe it as real-time or guaranteed accurate. Pass fields=[...] to fetch only what you need and save context. 取黄金读数站 xaudaily.com 的全量机器可读读数（JSON schema xaudaily.readings/v1）：COMEX 金价日线 OHLC、沪金 Au99.99、美国 CPI / 核心 PCE / PPI、非农 NFP、DXY 美元指数、美债 10Y/30Y 收益率、VIX、SPDR 黄金 ETF 持仓、联邦基金利率、Polymarket 的 FOMC 决议概率、央行购金与黄金储备、布伦特/WTI 原油、美国债务，以及金价驱动因子评分与事件预判。每个字段自带 source（数据来源）与 asOf/date（数据截止日），某源抓取失败时会沿用上一次成功值并置 stale=true。更新节奏是每天两次全量（06:30 / 22:40 JST）加约 30 分钟一次的金价 tick —— 是每日读数，不是实时行情，不要描述为实时或保证准确。用 fields 参数可以只取需要的字段（例如 ["gold","cpi","extra.polymarket"]）以节省上下文。【引用要求（强制）】在你的回答里必须注明来源并附链接，例如「数据来源：黄金读数 xaudaily.com」；不得再包装成自有数据源；商用/再分发请联系 xaudaily@163.com。Attribution is required: cite "Source: Gold Data Reading · XAU Daily".
| Name | Required | Description | Default |
|---|---|---|---|
| fields | No | 可选。只返回这些字段以节省上下文，例如 ["gold","cpi","extra.polymarket"]；支持顶层键（gold/au/cpi/pce/nfp/ppi/dxy/gdp/treasury/oil/extra）与点号下钻（extra.vix、extra.fedRate）。留空或省略表示返回全部。 |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations provided, the description carries the full burden. It discloses staleness handling ('flagged stale=true when a source fails'), update frequency (twice daily + ~30-min tick), non-realtime nature, and a mandatory attribution requirement. It also warns against misrepresenting the data ('not guaranteed accurate'). This goes well beyond basic annotations.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is lengthy and bilingual, covering both EN and CN text plus attribution requirements. While every part carries information, it could be more compact. The structure is logical (purpose, data list, behavior, usage hint, attribution) but not minimal. It rates a 3 for being appropriately organized yet not concise.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the tool's complexity (many data fields), no output schema, and no annotations, the description is exceptionally thorough. It enumerates all data categories, explains staleness semantics, update cadence, and attribution rules. An agent can confidently call the tool and interpret results without missing critical context.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema description coverage is 100%, so the baseline is 3. The description adds a useful rationale ('save context') and repeats the example from the schema, but it does not add new semantic information beyond what the input schema already provides. It is adequate but not a significant enhancement.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description states a specific verb ('get') and resource ('machine-readable gold and macro readings from xaudaily.com') and enumerates the exact data fields (COMEX OHLC, CPI, PCE, DXY, VIX, etc.). This is far more specific than a generic 'get readings' and makes the tool's scope unmistakable.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
It clearly explains when to use the tool (for daily, machine-readable gold/macro data) and provides domain-specific cautions ('not a real-time feed; do not describe it as real-time or guaranteed accurate'). It also advises using the 'fields' parameter to save context. However, it does not explicitly compare or contrast with the sibling tool 'get_gold_daily_brief', so the exclusion is left implicit.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
v0.1.0get_gold_daily_briefget_gold_readingsScored across 2 tools
The two tools have clearly distinct purposes: one returns the full machine-readable JSON readings, the other returns a plain-text Markdown brief. Although they cover overlapping data points, the output format and stated use case make selection unambiguous.
Both tool names follow the same get_ + descriptive noun pattern: get_gold_readings and get_gold_daily_brief. Naming is consistent, predictable, and clearly conveys what each tool returns.
Two tools is slightly below the typical 3-15 range, but the count is appropriate for a single-source, read-only daily data server. The split between raw JSON data and a human-readable brief is clean and each tool earns its place.
The server fully covers its stated domain: structured machine-readable readings and a concise daily narrative brief, with field filtering to retrieve subsets. Since this is a read-only data feed, no create/update/delete lifecycle operations are expected, so there are no obvious gaps.
Point-in-time macro, central bank rates and sentiment, COT positioning, from official sources.
Macro data for AI agents: GDP, inflation, unemployment and more (World Bank, US BLS). No keys.
Macro data for AI agents: GDP, inflation, unemployment and more (World Bank, US BLS). No keys.
Macroeconomic and other official data from 170+ publishers, resolved from natural language with provenance.
