Keyword Research MCP: autocomplete, search volume, difficulty and Google Trends

Keyword Research MCP: autocomplete, search volume, keyword difficulty and Google Trends for AI agents

A remote MCP server with six read-only keyword research tools that Claude, Cursor, VS Code and other MCP clients and AI agents can call: Google, YouTube and Bing autocomplete; keyword search volume, CPC, competition, keyword difficulty and search intent; keyword ideas; and Google Trends (interest over time, related queries, trending searches).

It runs on Apify as the Actor fairfetch/keyword-research-mcp, published by fairfetch, an independent developer on Apify Store. You pay per call through your Apify account. There is no subscription, and you don't need an Ahrefs, Semrush, Google Ads or DataForSEO account.

This repository holds only the listing: this README, server.json for the MCP Registry, and the icon. The server's source code is not published.

Tools and prices

Prices as read from the Apify API on 27 Oct 2026; the pricing on the Actor's Apify page is authoritative. Failed calls, invalid arguments and calls without data cost nothing, and keywords without data are not counted. Arguments, output fields and examples of every tool are documented on the Actor's Apify page.

Apify platform usage, on top of these prices. The server runs in Apify's Standby mode, and for Standby Actors Apify bills your account for the server's platform usage:

• The server starts when your MCP client connects to it, and most clients connect when they start or when a chat starts, even if no tool is used. It stops 5 minutes after the last request. Each start came to about $0.005 at 256 MB in our tests, the 5-minute tail included.
• Each start also costs Apify's start fee of $0.00005 and one small request through Apify's residential proxy that prepares Google Trends (about $0.0001).
• Every Google Trends call adds about $0.00015 of proxy traffic.
• If your client sends more than 8 requests at the same moment, Apify starts a second server, which adds its own 5-minute tail (about $0.0045).

Cost control: read this before giving an agent these tools

Each call is bounded by the tool's own argument limits. The most a single call can cost at today's prices:

An agent that keeps calling is not stopped by these limits. What limits its spending:

• Every tool description includes its price, so the agent sees what a call costs, and an answer is billed only when it has data.
• Every run has a maximum charge. For this MCP server Apify sets it itself (in our test it matched the account's remaining usage limit). A call that would go over it is refused before anything is fetched, and the error says how much is left.
• Set your Apify usage limit to what you are willing to spend, in Apify Console under Billing > Limits (Custom usage limit), up to your plan's maximum. Apify suspends the account's services when the limit is reached. On the Free plan the limit is your monthly free usage (the Free plan includes $5 of usage per month, pricing); to allow more, upgrade or wait for the next billing cycle.
• A running server keeps the maximum it started with. After raising the limit, close or disconnect your MCP client for 5 minutes, so that a new server starts with the new limit.
• Remove or disable the server in clients and projects where you don't use it. A connected client can start the server (about $0.005 each time) without calling any tool, for example each time the client starts or opens a new chat.

Connect

Option 1: Apify's hosted MCP server (sign in with Apify, no token to copy)

Add this URL as a remote MCP server (Streamable HTTP). Your client then asks you to sign in to Apify (OAuth) in your browser:

https://mcp.apify.com/?tools=fairfetch/keyword-research-mcp

Claude Code

claude mcp add --transport http keyword-research "https://mcp.apify.com/?tools=fairfetch/keyword-research-mcp"

Then run /mcp in Claude Code, pick keyword-research and sign in to Apify (or run claude mcp login keyword-research).

Claude apps: Settings > Connectors > Add custom connector, then paste the URL.

Cursor (~/.cursor/mcp.json, or .cursor/mcp.json in a project):

{  "mcpServers": {    "keyword-research": {      "url": "https://mcp.apify.com/?tools=fairfetch/keyword-research-mcp"    }  }}

VS Code: run MCP: Add Server, choose HTTP and paste the URL, or add to .vscode/mcp.json:

{  "servers": {    "keyword-research": {      "type": "http",      "url": "https://mcp.apify.com/?tools=fairfetch/keyword-research-mcp"    }  }}

Through this URL the tools appear as fairfetch--keyword-research-mcp--search_suggestions and so on. Apify's OAuth sign-in gives the client access to your Apify account (Apify's OAuth offers a single scope, full API access). Clients without a browser can send the header Authorization: Bearer <your Apify API token> to this URL instead.

Option 2: the server's own endpoint, with an Apify API token

For headless agents and SDKs. This endpoint accepts an Apify API token only (no OAuth sign-in). Get a token in Apify Console > Settings > API & Integrations and send it as Authorization: Bearer <token>:

https://fairfetch--keyword-research-mcp.apify.actor/mcp

Claude Code

claude mcp add --transport http keyword-research https://fairfetch--keyword-research-mcp.apify.actor/mcp \  --header "Authorization: Bearer YOUR_APIFY_TOKEN"

Cursor:

{  "mcpServers": {    "keyword-research": {      "url": "https://fairfetch--keyword-research-mcp.apify.actor/mcp",      "headers": { "Authorization": "Bearer YOUR_APIFY_TOKEN" }    }  }}

VS Code (.vscode/mcp.json; VS Code asks for the token and stores it securely):

{  "inputs": [    { "type": "promptString", "id": "apify-token", "description": "Apify API token", "password": true }  ],  "servers": {    "keyword-research": {      "type": "http",      "url": "https://fairfetch--keyword-research-mcp.apify.actor/mcp",      "headers": { "Authorization": "Bearer ${input:apify-token}" }    }  }}

Here the tool names have no prefix (search_suggestions ...), and a misspelled argument gets a free error that names it. Apify's hosted server (option 1) drops arguments a tool doesn't declare before it calls this server.

Data sources

Keyword metrics and ideas cover 94 countries (not China or Russia); autocomplete and Google Trends work for any country (Bing: 34 markets).

Legal

The tools read public search suggestions and Google Trends statistics, and keyword statistics retrieved through the DataForSEO API. They do not log in anywhere or solve CAPTCHAs, and they collect no personal data. Suggestions and trending searches are generated from what many people search for and can include names of public figures and, rarely, other text you should handle with care.

How the Google Trends tools reach Google. Their requests go through Apify's residential proxy (Apify's datacenter proxy if residential isn't available) with a Chrome browser User-Agent, one at a time per IP address and about a second apart. When Google throttles a request (HTTP 429 or 403, or a redirect to its "unusual traffic" page), the request is retried from another IP address, up to 5 times. If neither proxy is available, the requests come from the server's own IP address instead, about 5 seconds apart, and a throttled request waits up to about a minute before each retry. Each IP address is also replaced after 30 requests, so that no single address carries a long series of requests. Google's cookies are kept when the IP address changes, and a connection whose cookie is still new can be given a cookie that Google has already accepted on another of the tools' connections. trends.google.com's robots.txt disallows only its /explore pages (/explore? and /trends/explore?), not the endpoints these tools use (/trends/api/..., /_/TrendsUi/data/batchexecute and /trending/rss).

The services' terms restrict automated access and reuse:

• Google and YouTube (suggestqueries.google.com, which has no robots.txt; trends.google.com): Google's Terms of Service prohibit automated access that ignores machine-readable instructions such as robots.txt, and YouTube's Terms of Service prohibit automated access without YouTube's prior written permission (except for public search engines).
• Bing (api.bing.com, whose robots.txt does not load): the Microsoft Services Agreement says the text and other material available on Bing and MSN is for noncommercial, personal use only, and that using it to build your own products is allowed only as far as Microsoft or the rights holders specifically authorize (section 14.f.i, "Bing and MSN Materials"). It also prohibits circumventing restrictions on access to its services, including "impermissible scraping" (section 3.a.vi).

Amazon, eBay and DuckDuckGo autocomplete are not offered: their robots.txt or terms exclude automated or AI-agent access. You are responsible for how you use the data and for complying with the terms of these services and the laws that apply to you.

Not affiliated with or endorsed by Google, YouTube, Microsoft/Bing, DataForSEO or any other company or product named here. Names are trademarks of their owners and are used only to describe the data sources and compatible software.

Support

Questions, bugs and feature requests: the Issues tab on the Actor's Apify page. This repository only holds the listing, so its own issue tracker is turned off.

MIT licensed (this repository contains only documentation, configuration and the icon).