Google Search Trends MCP 实战：让 AI 帮你在趋势爆发前抓住它

Google Search Trends MCP: Let Your AI Catch Trends Before They Peak

Jul 6, 2026

更新于 Jul 6, 2026

🇨🇳 中文

MCP 页面: apify.com/trendsmcp/google-search-trends-mcp 服务地址: https://api.trendsmcp.ai/mcp · 免费 Key: trendsmcp.ai（100 次/月，免信用卡） 数据源: Google 搜索/图片/新闻/购物、YouTube、TikTok、Reddit、Amazon、Wikipedia、npm、Steam、App 榜单…共 25+

一、这个 MCP 到底解决什么问题

做内容、做产品、做投放的人都知道一句老话：“你不是没有好点子，你只是比别人晚了 60 天。”

问题出在工具链上。传统 SEO 工具（基于点击流数据）看到一个趋势时，往往已经滞后真实曲线 30 到 90 天——等你看到那条漂亮的上升曲线，红利期已经过了，曲线正在变平。而真正赚到钱、做出爆款的人，是在曲线刚开始往上翘的那一刻就进场的。

Google Search Trends MCP 想解决的就是这个”信息时差”问题。它把 25+ 个数据源的趋势数据，通过 MCP 协议直接喂给你的 AI（Claude、Cursor、任何支持 MCP 的客户端），让你可以用自然语言直接问：

“帮我查一下 ‘AI 陪伴玩具’ 最近 3 个月的搜索趋势，再看看 Reddit 和 TikTok 上是不是同步在涨。”

它有三个核心特点，恰好戳中普通用户的痛点：

不用爬虫、不用 Google API Key——你不需要懂技术、不需要申请任何官方接口，注册拿个 Key 就能用。
25+ 数据源交叉验证——不只是 Google，还有 YouTube、TikTok、Reddit、Amazon、Wikipedia、npm、Steam、App 下载榜。一个趋势是真火还是假火，交叉一比就知道。
AI 原生——数据不是丢给你一张图让你自己看，而是直接进 AI 的上下文，AI 能帮你分析、对比、写文案、列选题。

二、三个工具，一次讲清

这个 MCP 只暴露三个工具，简单到不能再简单：

工具	作用	你会怎么用
`get_trends`	任意关键词过去约 5 年的周级时间序列，归一化到 0–100	”这个词是长期在涨，还是一波流？“
`get_growth`	指定周期（7天/1月/3月/6月/1年/年初至今）的增长百分比	”过去 30 天涨了多少？是不是 breakout（爆发）？“
`get_top_trends`	实时热榜，不用输关键词，直接给你现在什么在涨	”现在全网最热的是什么？我该蹭哪个？”

关键概念——Breakout（爆发）：当一个词在给定周期内增长超过 5000%，就会被标记为 Breakout。这类词的特点是：搜索意图强、竞争小、绝大多数人还没反应过来。这就是你要找的金矿。

⚠️ 一个诚实的提醒：这个服务返回的数值是”归一化的专有估算值”，不是官方平台的真实指标。它由社区开发者（TrendsMCP）运营，适合判断方向和相对趋势，不适合当作精确的绝对数据去写财报。把它当成”趋势雷达”，而不是”官方统计局”。

三、5 分钟接入（普通用户版）

第 0 步：拿一个免费 Key

打开 trendsmcp.ai，填邮箱，秒收 API Key。免费额度 100 次/月，不用信用卡。够个人用户先玩一阵了。

价格阶梯（按月请求数）：

套餐	每月请求	价格
Free	100	$0
Starter	1,000	$19
Pro	5,000	$49
Business	25,000	$199

第 1 步：把它塞进你的客户端

Claude Desktop（最常见）——编辑 claude_desktop_config.json，加一段：

{
  "mcpServers": {
    "trends-mcp": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://api.trendsmcp.ai/mcp",
               "--header", "Authorization:${AUTH_HEADER}"],
      "env": { "AUTH_HEADER": "Bearer 你的_TRENDS_API_KEY" }
    }
  }
}

Cursor / Windsurf / VS Code——它们支持直连 HTTP transport，更简单：

{
  "trends-mcp": {
    "url": "https://api.trendsmcp.ai/mcp",
    "transport": "http",
    "headers": { "Authorization": "Bearer 你的_TRENDS_API_KEY" }
  }
}

💡 找不到配置文件在哪？Claude Desktop 里点 设置 → Developer → Edit Config 就能定位。改完完全退出再重开客户端，别只是关窗口。

第 2 步：验证

重启后，直接对 AI 说：

“用 trends-mcp 查一下 ‘agentic AI’ 过去一年的搜索趋势。”

如果它调用了工具并返回一串 0–100 的数值，就成了。

四、真正的价值：从”查数据”到”抓机会”

接入只是入场券。真正拉开差距的，是你怎么问、怎么串起来分析。下面这套方法论，是把 Trends 从”玩具”变成”生产力”的关键。

方法 1：读懂曲线的”第一导数”——抢在爆发前进场

不要只看”现在热不热”，要看”上升的速度”。一条趋势曲线有四个阶段：

        ╱‾‾‾╲
   ②   ╱     ╲  ③
      ╱       ╲___
① ___╱            ④
  沉睡   起飞    见顶   衰退

① 沉睡：搜索量低、平；大多数人不知道。
② 起飞（第一导数为正，且在变大）：曲线开始指数上翘。**这就是你要进场的点。**竞争小，算法正饥渴地寻找优质内容源。
③ 见顶：曲线变平。传统工具现在才告诉你——太晚了。
④ 衰退：从峰值跌超 20%，别碰。

给 AI 的提问模板：

“用 get_trends 查这批关键词过去 6 个月的曲线：[A, B, C]。帮我判断每一个现在处于沉睡/起飞/见顶/衰退哪个阶段，只把处于『起飞』阶段的挑出来。”

再叠加 get_growth 做量化确认：

“对上面挑出来的词，用 get_growth 查 7 天、1 个月、3 个月的增长率。如果 7 天 > 1 个月 > 0 且加速，说明还在起飞，标为⭐。“

方法 2：交叉验证——区分”真趋势”和”一波流”

单一数据源会骗人。一个词在 Google 上涨，可能只是一条新闻带起来的 48 小时热点；但如果它在 Google 搜索 + YouTube + Reddit + Amazon 上同步在涨，那大概率是真需求在形成。

判断标准：

一波流（news-cycle）：单一来源尖峰，48 小时内涨落 → 只适合做时效性快内容，别投产品。
真趋势（evergreen-with-a-spike）：多来源同步、持续爬升数周 → 值得下重注（选品、建站、做系列内容）。

给 AI 的提问模板：

“‘便携榨汁杯’ 这个词，帮我在 Google 搜索、YouTube、TikTok、Amazon、Reddit 五个源上分别查趋势。如果多数源同步上涨，判定为真需求；如果只有一个源在涨，判定为一波流。给出结论和理由。“

方法 3：内容创作——用”热榜 + 派生词”批量产选题

get_top_trends 给你实时热榜，但热榜大家都看得到，竞争激烈。真正的选题金矿在**派生词（rising related queries）**里——热门大词下面那些正在快速上涨的长尾。

工作流：

get_top_trends 拉当前热榜 → 选一个和你领域相关的大词。
让 AI 围绕这个大词发散出具体的、正在上涨的子问题（谁、为什么、怎么做、对比、平替）。
对每个子问题用 get_trends 确认还在起飞。
24–48 小时内发布——趋势内容的黄金窗口。晚了就成红海。

给 AI 的提问模板：

“用 get_top_trends 拉现在的热榜，挑出和『AI 工具』相关的话题。针对每个话题，帮我列 5 个正在上涨的长尾选题（要具体到能直接写成一篇文章的标题），并按『我能在 48 小时内发布 / 需要 2 周深做』分两类。“

方法 4：商机捕捉——找”窄而深”的 Breakout 缝隙

赚钱的机会往往不在大词里（大词早被巨头占了），而在类目里那些刚爆发、还没人做的窄缝。

工作流：

锁定一个你熟悉的类目（不是宽泛大词）。
让 AI 在这个类目下找 Breakout（增长 >5000%）或 Rising 的具体词。
用多源交叉验证过滤掉一波流（方法 2）。
对活下来的词，评估：能做成什么？（选品 / 工具 / 内容站 / 服务）

给 AI 的提问模板：

“我想在『家庭健身』这个类目里找商机。帮我用 trends-mcp 找出这个类目下过去 3 个月增长最快的 10 个细分词，过滤掉只有单一数据源在涨的（一波流），剩下的每一个告诉我：搜索意图是什么、目前有没有明显的头部玩家、如果我要切进去最轻的切入方式是什么（内容 / 选品 / 工具）。“

方法 5：灵感发现——跨源”串联”出没人讲过的角度

最高级的用法：把不相关的上升趋势串起来，找到”叙事缝隙（narrative gap）“。当两个原本不搭界的趋势同时在涨，它们的交集往往是一个全新的、还没人占据的内容/产品角度。

给 AI 的提问模板：

“用 get_top_trends 分别拉『科技』和『健康』两个类目的上升词。帮我找出可以交叉的组合（比如某个科技趋势 × 某个健康趋势），每个组合给我一个还没什么人做、但需求正在形成的内容或产品点子。“

五、把它变成一条自动化流水线

上面五个方法，最终可以固化成一个每天自动跑的 Agent 流程（配合 Claude Code 的定时任务或任何 Agent 框架）：

每天早上：
  1. get_top_trends 拉全网热榜
  2. 过滤出我的 3 个关注类目
  3. 每个类目找 Rising/Breakout 词
  4. 多源交叉验证，剔除一波流
  5. 幸存的词 → 生成选题清单 + 商机评估
  6. 标记『48h 可发』的，直接起草初稿
  → 一份『今日趋势机会简报』推到我手机

这就是 MCP 的真正威力：它不是给你一个查询框，而是给你的 AI 装上了一双”看得比别人早 60 天”的眼睛。

六、几个诚实的边界

数据是估算值：适合判断方向和相对强弱，不适合当精确绝对值。
免费额度有限：100 次/月，方法 2（多源交叉）一次就要好几个请求，重度用户需要升级。
趋势 ≠ 变现：Trends 告诉你”需求在涨”，但能不能接住需求，取决于你的执行。工具负责发现，你负责落地。
社区服务：由独立开发者运营，不是 Google 官方，稳定性和长期性需自行评估。

但即便有这些边界，对绝大多数内容创作者、独立开发者、小团队来说，一个能让 AI 主动、跨源、实时读懂趋势的入口，已经是一个不小的杠杆。趋势的红利从来不奖励看得清的人，只奖励看得早的人。

参考资料

Google Search Trends MCP（Apify）: https://apify.com/trendsmcp/google-search-trends-mcp
TrendsMCP 官网与文档: https://trendsmcp.ai
Google Trends For SEO In 2026: The Velocity Playbook (Yotpo)
The 5,000 Percent Trick: How to Discover Breakout Trends (Xpert.digital)
How to use Google Trends for SEO in 2026 (Semrush)
7 Ways to Find Trending Topics Before They Peak (vidIQ)

🇬🇧 English

MCP page: apify.com/trendsmcp/google-search-trends-mcp Endpoint: https://api.trendsmcp.ai/mcp · Free key: trendsmcp.ai (100 req/mo, no credit card) Sources: Google Search/Images/News/Shopping, YouTube, TikTok, Reddit, Amazon, Wikipedia, npm, Steam, App charts… 25+ total

1. What problem does this MCP actually solve

Anyone in content, products, or ads knows the old line: “You’re not short on good ideas — you’re just 60 days too late.”

The blame lies with the tooling. Traditional SEO tools (built on clickstream data) surface a trend 30 to 90 days after the real curve — by the time you see that pretty upward line, the window has closed and the curve is already flattening. The people who actually make money and produce hits enter the moment the curve starts to bend upward.

Google Search Trends MCP exists to kill that time lag. It pipes trend data from 25+ sources through the MCP protocol straight into your AI (Claude, Cursor, any MCP-capable client), so you can ask in plain language:

“Pull the last 3 months of search trend for ‘AI companion toys’, then check whether Reddit and TikTok are rising in sync.”

Three traits make it matter for normal users:

No scraping, no Google API key — you don’t need to be technical or apply for any official API. Grab a key and go.
25+ sources for cross-validation — not just Google, but YouTube, TikTok, Reddit, Amazon, Wikipedia, npm, Steam, App download charts. Real trend or fake spike? Cross-check and you know.
AI-native — the data doesn’t dump a chart for you to eyeball; it enters the AI’s context so it can analyze, compare, draft copy, and generate topic lists for you.

2. Three tools, explained once

The MCP exposes just three tools — dead simple:

Tool	What it does	How you’ll use it
`get_trends`	~5 years of weekly time-series for any keyword, normalized 0–100	”Long-term climb or one-hit spike?”
`get_growth`	Percentage change over a period (7D / 1M / 3M / 6M / 1Y / YTD)	“How much did it grow in 30 days? Is it a breakout?”
`get_top_trends`	Live leaderboard, no keyword needed — what’s rising right now	”What’s hot everywhere right now? Which wave do I ride?”

Key concept — Breakout: when a term grows over 5,000% in a period, it’s flagged as Breakout. These have strong intent, low competition, and almost nobody has noticed yet. That’s the gold you’re hunting.

⚠️ An honest caveat: the values returned are “normalized proprietary estimates”, not official platform metrics. It’s run by a community developer (TrendsMCP), best for judging direction and relative trend, not for citing precise absolute numbers in a financial report. Treat it as a “trend radar”, not a “national statistics bureau”.

3. Five-minute setup (normal-user edition)

Step 0: Get a free key

Open trendsmcp.ai, enter your email, receive the API key instantly. Free tier: 100 req/mo, no credit card.

Pricing tiers (by monthly requests): Free 100 = $0 · Starter 1,000 = $19 · Pro 5,000 = $49 · Business 25,000 = $199.

Step 1: Wire it into your client

Claude Desktop — edit claude_desktop_config.json:

{
  "mcpServers": {
    "trends-mcp": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://api.trendsmcp.ai/mcp",
               "--header", "Authorization:${AUTH_HEADER}"],
      "env": { "AUTH_HEADER": "Bearer YOUR_TRENDS_API_KEY" }
    }
  }
}

Cursor / Windsurf / VS Code — direct HTTP transport, simpler:

{
  "trends-mcp": {
    "url": "https://api.trendsmcp.ai/mcp",
    "transport": "http",
    "headers": { "Authorization": "Bearer YOUR_TRENDS_API_KEY" }
  }
}

💡 Can’t find the config file? In Claude Desktop: Settings → Developer → Edit Config. After editing, fully quit and reopen the app — don’t just close the window.

Step 2: Verify

After restart, tell your AI: “Use trends-mcp to pull the last year of search trend for ‘agentic AI’.” If it calls the tool and returns 0–100 values, you’re live.

4. The real value: from “querying data” to “catching opportunity”

Setup is just the ticket in. What separates people is how you ask and how you chain the analysis. This methodology turns Trends from a toy into a production tool.

Method 1: Read the “first derivative” — enter before the breakout

Don’t just look at “is it hot now” — look at the speed of the climb. A trend curve has four stages: ① Dormant (low, flat) → ② Lift-off (first derivative positive and growing — this is your entry point, low competition, algorithm hungry for good sources) → ③ Peak (flattening — traditional tools only tell you now, too late) → ④ Decline (down >20% from peak — don’t touch).

Prompt template:

“Use get_trends on these keywords over the last 6 months: [A, B, C]. Classify each as dormant/lift-off/peak/decline, and surface only the ones in ‘lift-off’.”

Then confirm quantitatively with get_growth:

“For those, pull 7-day, 1-month, 3-month growth. If 7D > 1M > 0 and accelerating, it’s still lifting off — flag it ⭐.”

Method 2: Cross-validate — separate real trends from one-hit spikes

A single source lies. A term rising on Google might just be a 48-hour news blip. But if it rises in sync across Google + YouTube + Reddit + Amazon, real demand is likely forming.

News-cycle spike: single-source peak, rises and falls in 48h → good only for timely fast content, don’t invest in a product.
Real trend (evergreen-with-a-spike): multi-source, sustained climb over weeks → worth a heavy bet (product selection, site building, content series).

Prompt template:

“For ‘portable juicer cup’, pull trends across Google Search, YouTube, TikTok, Amazon, and Reddit. If most sources rise in sync, judge it real demand; if only one rises, judge it a one-hit spike. Give the conclusion and reasoning.”

Method 3: Content creation — mass-produce topics from “leaderboard + derived queries”

get_top_trends gives the live board, but everyone sees the board — it’s crowded. The real gold is in rising related queries — the long-tail climbing fast underneath the big terms.

Workflow: pull get_top_trends → pick a big term in your niche → have the AI branch out into specific, rising sub-questions (who / why / how / vs / cheaper alternative) → confirm each is still lifting via get_trends → publish within 24–48 hours (the golden window; later = red ocean).

Prompt template:

“Use get_top_trends, pick topics related to ‘AI tools’. For each, list 5 rising long-tail topics (specific enough to be an article headline), split into ‘I can publish in 48h’ vs ‘needs 2 weeks deep work’.”

Method 4: Opportunity capture — find the “narrow and deep” breakout gaps

Money is rarely in the big terms (giants own those) — it’s in the narrow gaps inside a category that just broke out and nobody’s serving yet.

Workflow: lock a category you know (not a broad term) → have the AI find Breakout (>5,000%) or Rising specific terms in it → filter one-hit spikes via Method 2 → for survivors, assess: what can this become? (product / tool / content site / service)

Prompt template:

“I want opportunities in ‘home fitness’. Use trends-mcp to find the 10 fastest-growing sub-terms in this category over the last 3 months, filter out single-source spikes, and for each survivor tell me: the search intent, whether there’s an obvious incumbent, and the lightest way to enter (content / product / tool).”

Method 5: Inspiration discovery — chain sources into an angle nobody’s taken

The advanced move: chain unrelated rising trends to find the “narrative gap”. When two unrelated trends rise at once, their intersection is often a fresh, unclaimed content/product angle.

Prompt template:

“Use get_top_trends to pull rising terms in both ‘tech’ and ‘health’. Find crossable combinations (a tech trend × a health trend), and for each give me a content or product idea nobody’s really doing yet but where demand is forming.”

5. Turn it into an automated pipeline

The five methods can be frozen into a daily agent run (with Claude Code’s scheduled tasks or any agent framework):

Every morning:
  1. get_top_trends → global leaderboard
  2. filter to my 3 watched categories
  3. find Rising/Breakout terms per category
  4. cross-validate, drop one-hit spikes
  5. survivors → topic list + opportunity assessment
  6. flag "publishable in 48h", auto-draft those
  → a "Today's Trend Opportunities" brief to my phone

That’s the real power of MCP: it’s not a search box — it gives your AI a pair of eyes that see 60 days earlier than everyone else.

6. Honest boundaries

Data is estimated — good for direction and relative strength, not precise absolutes.
Free tier is limited — 100/mo; Method 2 (multi-source) burns several requests each time. Heavy users must upgrade.
Trend ≠ monetization — Trends tells you demand is rising; whether you catch it depends on execution. The tool discovers; you deliver.
Community service — run by an independent developer, not Google. Assess stability and longevity yourself.

Even so, for most creators, indie developers, and small teams, an entrance that lets your AI read trends proactively, cross-source, in real time is real leverage. Trend dividends never reward those who see clearly — only those who see early.

References

Google Search Trends MCP (Apify): https://apify.com/trendsmcp/google-search-trends-mcp
TrendsMCP site & docs: https://trendsmcp.ai
Google Trends For SEO In 2026: The Velocity Playbook (Yotpo)
The 5,000 Percent Trick: How to Discover Breakout Trends (Xpert.digital)
How to use Google Trends for SEO in 2026 (Semrush)
7 Ways to Find Trending Topics Before They Peak (vidIQ)

© 2026 Author: Mycelium Protocol. Licensed under CC BY 4.0.

💬 评论与讨论

使用 GitHub 账号登录后发表评论

Google Search Trends MCP 实战：让 AI 帮你在趋势爆发前抓住它

Google Search Trends MCP: Let Your AI Catch Trends Before They Peak

一、这个 MCP 到底解决什么问题

二、三个工具，一次讲清

三、5 分钟接入（普通用户版）

第 0 步：拿一个免费 Key

第 1 步：把它塞进你的客户端

第 2 步：验证

四、真正的价值：从”查数据”到”抓机会”

方法 1：读懂曲线的”第一导数”——抢在爆发前进场

方法 2：交叉验证——区分”真趋势”和”一波流”

方法 3：内容创作——用”热榜 + 派生词”批量产选题

方法 4：商机捕捉——找”窄而深”的 Breakout 缝隙

方法 5：灵感发现——跨源”串联”出没人讲过的角度

五、把它变成一条自动化流水线

六、几个诚实的边界

参考资料

1. What problem does this MCP actually solve

2. Three tools, explained once

3. Five-minute setup (normal-user edition)

Step 0: Get a free key

Step 1: Wire it into your client

Step 2: Verify

4. The real value: from “querying data” to “catching opportunity”

Method 1: Read the “first derivative” — enter before the breakout

Method 2: Cross-validate — separate real trends from one-hit spikes

Method 3: Content creation — mass-produce topics from “leaderboard + derived queries”

Method 4: Opportunity capture — find the “narrow and deep” breakout gaps

Method 5: Inspiration discovery — chain sources into an angle nobody’s taken

5. Turn it into an automated pipeline

6. Honest boundaries

References

相关文章

💬 评论与讨论