A Google Trends MCP server is a connector that lets Claude, ChatGPT or Cursor read Google Trends data when you ask. MCPifex hosts one with five read-only tools: interest over time, term comparison, related queries, trending now and interest by region. It needs no Google account and no API key from Google, only an MCPifex key.

MCP, the Model Context Protocol, is how an AI app talks to outside tools. An MCP server is one tool the AI can call. This one wraps Google Trends. Instead of opening trends.google.com, typing a term, switching the date range and copying numbers into a sheet, you ask in plain English and the AI makes the calls and explains the result. If MCP is new to you, start with what MCP is.

The setup is the same in ChatGPT or Claude: an AI chat that can fetch live Trends numbers instead of guessing from training data. That matters because a model without a data source will happily make up a trend. With the server connected, the numbers in the answer came from a tool call, and you can ask to see it.

Which tools does it have?

There are five tools. All are read-only and all are on by default, so there is nothing risky to switch off.

ToolWhat it doesA good use
interest_over_timeInterest scores (0 to 100) for up to 5 terms over a time range, one series per term.Is this topic growing, flat or seasonal?
compare_termsCompares 2 to 5 terms, with per-term averages and the highest-interest term.Which name or keyword wins?
related_queriesRising and top related searches for a term.Find adjacent topics and new angles.
trending_nowToday's trending searches for a country, with estimated traffic and linked articles.Spot news and fast movers.
interest_by_regionInterest for a term broken down by region, showing the top regions.Pick a market or ad geography.

Does it need an API key or a Google login?

No. The Google Trends server is keyless. It needs no form fields or sign-in. In the portal you name the instance, leave the tools on, and generate an MCPifex API key. The quickstart shows the portal steps. The Free plan includes 3 instances and needs no card, so this is a few minutes of work and costs nothing to try.

Your MCPifex key is the only credential. It lets your AI reach the gateway, and you can revoke it from the portal.

First create a free account, set up the instance and generate an API key. It starts with mcpx_. Then replace <YOUR_MCPX_KEY> below with your key.

Claude Code

#!/bin/sh
# MCPifex — Claude Code CLI snippet.
# Replace <YOUR_MCPX_KEY> with your MCPifex API key (starts "mcpx_"), from
# the portal's API key page, then run this once from any shell that has the
# `claude` CLI on PATH.
claude mcp add mcpifex --transport http https://mcp.mcpifex.com/mcp \
  --header "Authorization: Bearer <YOUR_MCPX_KEY>"

Cursor

Add this to .cursor/mcp.json in your project, or to Cursor's global mcp.json, then reload Cursor:

{
  "mcpServers": {
    "mcpifex": {
      "type": "http",
      "url": "https://mcp.mcpifex.com/mcp",
      "headers": {
        "Authorization": "Bearer <YOUR_MCPX_KEY>"
      }
    }
  }
}

Claude Desktop

Use Claude's remote custom connector where it's available. Otherwise, this mcp-remote bridge works, and needs Node.js and npx. Merge the mcpifex entry into the mcpServers object in claude_desktop_config.json, then restart the app:

{
  "mcpServers": {
    "mcpifex": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://mcp.mcpifex.com/mcp",
        "--header",
        "Authorization:${MCPIFEX_AUTH_HEADER}"
      ],
      "env": {
        "MCPIFEX_AUTH_HEADER": "Bearer <YOUR_MCPX_KEY>"
      }
    }
  }
}

ChatGPT

ChatGPT's custom MCP form has no header field, so the key goes in the URL: https://mcp.mcpifex.com/mcp/<YOUR_MCPX_KEY>. The quickstart has the step-by-step ChatGPT setup. Treat that URL like a password.

What can I ask it? Ten example prompts

Write these the way you would brief a colleague. Name the terms, the country and the time range, since the tools take those as inputs.

  1. "Compare interest in 'running shoes', 'trail shoes' and 'barefoot shoes' in the US over the past 5 years. Which is growing fastest?"
  2. "Show interest over time for 'pickleball' worldwide for 12 months. Does it have a seasonal peak?"
  3. "What are the top and rising related queries for 'meal prep'? Group them into themes."
  4. "What is trending in Germany today? Flag anything relevant to fitness or food."
  5. "Which US states search for 'heat pump' the most? Give me the top ten regions."
  6. "I'm choosing between 'AI notetaker' and 'meeting assistant' for a landing page headline. Compare them and tell me which has more interest."
  7. "Pull related queries for 'standing desk' and suggest five blog post titles from the rising ones."
  8. "Compare my brand name with two competitors over the past year and describe the gap in plain words."
  9. "Find out when interest in 'tax software' peaks each year so I can plan a campaign calendar."
  10. "Which UK regions search for 'wedding venues' most? Suggest where to focus local ads."

For a full workflow that turns Trends results into a keyword plan, read keyword research with the Google Trends MCP server. It covers the research method. This page covers the server itself.

Partly. For years Google offered no public Trends API, and tools relied on the same endpoints the website uses. Google has now published a Google Trends API in alpha. Its official page describes programmatic access to Trends data, a rolling window of the last 5 years, aggregation by day, week, month or year, and data that is not scaled from 0 to 100. Access is limited: Google is accepting applications from alpha testers and prioritizing developers with a specific use case who can give feedback.

So an official alpha exists, but it is not open to everyone and is built for developers with an application. The MCPifex server does not use it. This server reads the public Trends endpoints without credentials, so its data follows the Trends website, including the 0 to 100 scale. If you are accepted into the alpha, you would still need your own code or client to call it.

Scores are relative, not search counts

Google Trends divides each data point by the total searches for that place and time, then scales the results from 0 to 100. A score of 50 does not mean 50 searches. It means a share of searches relative to the peak in the range you asked for. Google also notes that low-volume terms can appear as 0, which reflects low volume rather than necessarily no searches. For absolute volumes, pair Trends with a keyword tool.

Data is sampled

Trends works from a sample of searches, and Google filters out some of them, such as repeated searches from one person in a short window. The sample is enough for direction and comparison. Treat small gaps between two terms as noise, and be careful with rarely searched terms, where Google says noise is most noticeable. Read Google's own explanation, how Trends data is adjusted, before you quote a number in a report.

Rate limits: Google can return 429

The server reads Google's public but unofficial Trends endpoints. When requests are heavy, Google can answer with HTTP 429, which means too many requests. The server retries once, then returns a clear error to your AI. The limit comes from Google, not from MCPifex, and calls through MCPifex are not capped. If you see it, wait a bit and ask again. Avoid prompts that fan out into dozens of terms at once. Because the endpoints are unofficial, Google can also change them without notice, and results are best-effort.

Ranges and ranking behaviour

Interest scores depend on the time range and region you pick. The same term can show different numbers across ranges, so keep the same range when you compare reports over time.

You have fair options, and the right one depends on where you work.

  • purahmanian/google-trends-mcp is an open-source Google Trends MCP package. It is free and keyless, and you run it on your own machine with Node. It shares the same unofficial endpoints and the same rate-limit exposure.
  • HasData Google Trends MCP is open source and lists Claude Desktop and Claude Code support, with regional interest and related queries. It runs on a HasData account, so check their current terms.
  • Trends MCP is a hosted connector aimed at Claude users. Check its site for current plans.
  • The official Google Trends API alpha, covered above, suits developers who want programmatic access and can apply.

Choose the open-source package if you are comfortable with a terminal and want to run it yourself. Choose a vendor-hosted option if you want its data backend. Choose MCPifex if you want nothing to install, one URL for several servers, and the ability to allow or deny each tool. Related reading: best MCP servers, and the Google Search Console server, which adds your own site's search data next to market-wide Trends.

Good to know

  • Each AI session gets its own isolated process.
  • Every call is logged in the portal.
  • Tool calls are unlimited. Plans only limit instances: Free 3, Pro 10, Team 30. See pricing.
  • New to connecting clients? See connect Claude.

Frequently asked questions

  • No Google key or account is needed. You only need an MCPifex API key to reach the gateway.

Sources

  1. Google Trends API (alpha), Google Search Central.
  2. How Trends data is adjusted, Google Trends Help.
  3. purahmanian/google-trends-mcp, an open-source alternative.
  4. Google Trends, the data source.

Run this server through MCPifex

Run hosted MCP servers behind one endpoint and choose exactly which tools your AI can use.

Start freeBrowse the marketplace

← Back to servers