How do you connect Claude to SEMrush?

Quick Answer: You can connect Claude to SEMrush using the SEMrush Model Context Protocol (MCP) server. Once configured, Claude pulls live SEMrush data directly inside your conversations, replacing manual tab-switching with natural language queries for domain overviews, keyword gaps, and backlink analysis.

Copying data out of SEMrush, pasting it into Claude, and waiting for analysis is slow. There is a faster way. By connecting Claude to SEMrush through the Model Context Protocol (MCP), you give Claude direct access to live SEMrush data. Ask it a question, get an answer backed by real numbers, without leaving your workflow.

This tutorial walks through the exact setup steps and shows you how to run three high-value SEO research tasks: domain overviews, keyword gap analysis, and backlink pulls. If you're evaluating where this fits into a broader SaaS growth strategy, it also helps to understand how specialist partners approach B2B SaaS SEO and content-led acquisition.

What Is MCP and Why Does It Matter for SEO?

MCP (Model Context Protocol) is an open standard that lets AI models like Claude connect to external tools and data sources in real time. Instead of Claude working only from its training data, MCP gives it live access to APIs, databases, and services you authorise.

For SEO, this means Claude stops guessing and starts fetching. When you ask "what keywords is this competitor ranking for?", Claude calls the SEMrush API, retrieves current data, and returns a structured answer in seconds.

SEMrush published its own MCP server specifically to support this workflow. It exposes core SEMrush API endpoints through a protocol Claude can read natively.

What You Need Before You Start

Before configuring anything, confirm you have the following:

• A Claude Pro subscription (custom MCP connectors require Pro or above)
• A SEMrush account with API access enabled (Guru plan or above includes API units)
• Your SEMrush API key (found under Profile > API > Generate API Key)
• Node.js installed on your machine (version 18 or higher recommended)
• Claude Desktop app installed (the browser version does not support MCP configuration)

If you are using Claude via API or inside Cursor or VS Code, the MCP setup process is nearly identical but the config file location differs. This guide covers Claude Desktop.

How to Connect Claude to SEMrush: Step-by-Step

Step 1: Locate or Create Your Claude Desktop Config File

Claude Desktop reads MCP server configurations from a JSON file on your machine.

On macOS:

~/Library/Application Support/Claude/claude_desktop_config.json

On Windows:

%APPDATA%\Claude\claude_desktop_config.json

If the file does not exist, create it. If it does exist, open it in any text editor.

Step 2: Add the SEMrush MCP Server Configuration

Paste the following into your config file, replacing YOUR_SEMRUSH_API_KEY with your actual key:

{
  "mcpServers": {
    "semrush": {
      "command": "npx",
      "args": [
        "-y",
        "@semrush/mcp-server"
      ],
      "env": {
        "SEMRUSH_API_KEY": "YOUR_SEMRUSH_API_KEY"
      }
    }
  }
}

If you already have other MCP servers configured, add the semrush block inside the existing mcpServers object rather than replacing it.

Step 3: Save the File and Restart Claude Desktop

Save the config file, then fully quit and relaunch Claude Desktop. A partial close does not reload the config.

Step 4: Verify the Connection

Once Claude Desktop restarts, open a new conversation and type:

"Can you confirm you have access to SEMrush tools?"

Claude should respond by listing the available SEMrush functions it can call. If it responds that it has no tools available, double-check the JSON syntax in your config file (a missing comma or bracket is the most common error) and confirm your API key is correct.

Three SEO Research Tasks You Can Run Immediately

Once connected, these three workflows deliver the most immediate value for B2B SaaS SEO research.

Task 1: Domain Overview

A domain overview gives you a snapshot of any site's organic performance: estimated traffic, keyword count, authority score, and traffic trend over time.

Prompt to use:

"Pull a domain overview for [competitor.com] using SEMrush. Give me their organic traffic estimate, number of ranking keywords, authority score, and the top 5 organic keywords driving their traffic."

Claude calls the SEMrush Domain Overview endpoint, returns the data in a structured format, and you can immediately ask follow-up questions like "how does this compare to [yourdomain.com]?" without re-entering SEMrush.

What this replaces: Opening SEMrush, navigating to Domain Overview, entering the domain, reading the dashboard, and manually noting figures before switching back to your doc.

Task 2: Keyword Gap Analysis

Keyword gap analysis surfaces keywords your competitors rank for that you do not. For B2B SaaS, this is one of the fastest ways to find content opportunities at scale.

Prompt to use:

"Run a keyword gap analysis between [yourdomain.com] and [competitor1.com] and [competitor2.com] using SEMrush. Show me the top 20 keywords where both competitors rank in the top 20 but we are not ranking at all."

Claude queries the SEMrush Keyword Gap endpoint, filters the results based on your criteria, and returns a prioritised list. You can then extend the prompt:

"Of those 20 keywords, which have a search volume above 500 and a keyword difficulty below 50?"

Claude filters the existing data without making a new API call, giving you a shortlist in seconds.

What this replaces: Manually running the Keyword Gap tool, exporting a CSV, opening it in a spreadsheet, and applying filters. For teams building an editorial roadmap from those gaps, this often overlaps with the work done by B2B SaaS content marketing agencies that turn keyword opportunities into pipeline-focused content plans.

Task 3: Backlink Analysis

Backlink data from SEMrush via Claude lets you audit a competitor's link profile or investigate a specific referring domain quickly.

Prompt to use:

"Using SEMrush, pull the top 20 referring domains for [competitor.com] by authority score. Include the anchor text used and whether the link is follow or nofollow."

For prospecting, extend this:

"Which of these referring domains are publications or industry blogs rather than directories or forums? Flag any with an authority score above 40."

Claude applies reasoning on top of the raw SEMrush data, giving you a qualified prospect list rather than a raw export.

What this replaces: Running the Backlinks report in SEMrush, exporting, and manually reviewing each domain for relevance.

Combining All Three Into a Single Research Session

The real productivity gain comes from chaining these tasks inside one Claude conversation. Because Claude holds the full context of the session, you can build on each result.

A practical sequence for a B2B SaaS competitor audit looks like this:

1. Run a domain overview on your top competitor
2. Ask Claude to identify which traffic categories (branded vs. non-branded) dominate their keyword mix
3. Run a keyword gap to find non-branded terms they rank for that you do not
4. Pull their backlinks and filter for domains covering the same topics as those gap keywords
5. Ask Claude to draft a one-page content and link-building brief based on everything it has found

The entire sequence takes under 15 minutes. The same workflow manually across SEMrush tabs, spreadsheets, and a separate document would take 90 minutes or more. If you want to pair this with distribution after the research phase, it can also support decisions around B2B SaaS social media agencies and other acquisition channels.

Common Setup Issues and How to Fix Them

Claude says it has no tools available after restart Check the JSON config file for syntax errors. Use a JSON validator (jsonlint.com works) to confirm the file is valid before restarting Claude again.

API calls return no data Confirm your SEMrush plan includes API access. Free and Pro plans have limited or no API units. Also check that your API key is active under SEMrush Profile > API.

Claude returns an error about rate limits SEMrush API units are consumed per request. If you hit your monthly limit, calls will fail. Check your API unit balance in your SEMrush account dashboard.

The npx command fails on first run The first run downloads the @semrush/mcp-server package. This requires an internet connection and may take 30-60 seconds. If it times out, run npx -y @semrush/mcp-server manually in your terminal to trigger the download separately.

FAQs

What Claude plan do I need to connect to SEMrush via MCP? You need Claude Pro at minimum. The MCP configuration feature, which allows custom server connections, is not available on the free tier. Claude Pro costs $20 per month and supports multiple MCP servers running simultaneously.

Does connecting Claude to SEMrush use my SEMrush API units? Yes. Every data request Claude makes through the MCP server consumes SEMrush API units from your account. Guru and Business plans include a monthly API unit allowance. Monitor your usage under Profile > API in your SEMrush account to avoid unexpected limits.

Can I connect Claude to SEMrush without coding experience? The setup requires editing a JSON config file and having Node.js installed, but no programming is needed beyond that. The JSON template above can be copied and pasted directly. If you are comfortable using a text editor and following file path instructions, you can complete the setup in under 10 minutes.

What SEMrush data can Claude actually access through MCP? The SEMrush MCP server exposes the most commonly used endpoints: domain overview, organic research, keyword gap, backlink analysis, keyword overview, and competitor discovery. It does not expose every SEMrush feature, but it covers the core data sets used in most SEO research workflows.

Is this the same as using Make or Albato to connect Claude and SEMrush? No. Tools like Make and Albato build automated workflows that trigger on schedules or events. The MCP approach is interactive: Claude fetches data on demand during a live conversation, applies reasoning, and responds in context. Both approaches have their place, but MCP is better suited to research and analysis tasks where you need to ask follow-up questions.