Semrush MCP Server: Connect Semrush to Claude, ChatGPT and Gemini

This information is correct as of September 2026; Growth Method has made every attempt to ensure the accuracy and reliability of the information provided.

Is there an official Semrush MCP server? Yes. Semrush MCP is a hosted Model Context Protocol server that brings Semrush’s SEO data, backlink data, keyword research, traffic and market estimates, and read-only project data into Claude, ChatGPT, Cursor, VS Code, Perplexity and other AI clients. You connect it once, sign in with your Semrush account (or an API key), and then ask for a domain overview, a keyword gap or a competitor’s traffic split in a normal sentence.

Semrush is one of the platforms Growth Method itself lists on its homepage integrations grid, alongside Google Analytics, Search Console and Webflow, all of which already have their own MCP explainer on this site. This one closes that gap: what the server actually covers, how its unit-based pricing works, how to set it up in each client, and how it stacks up against the community Semrush servers and against DataForSEO MCP, the other official SEO-data server in our MCP servers roundup.

What Semrush MCP is (and is not)

Semrush is a large, broad SEO and market-data platform, and its MCP server mirrors that breadth rather than picking one corner of it. Per Semrush’s own developer documentation, the server provides data from three places:

• The full SEO API — domain overview, organic and paid research, keyword research, backlinks, competitor research and more.
• The full Trends API — traffic and market data, audience insights and growth benchmarks, gated to your Trends API subscription level.
• The read-only methods of the Projects API (v3) — you can pull data out of your existing Semrush projects, but the MCP cannot create a project, change its settings or perform any other write action.

That last point is worth underlining because it is the whole shape of the server: everything it does is a read. There is no tool anywhere in Semrush MCP that changes your Semrush account, starts a crawl, or edits a tracked keyword list. If you want to manage projects, you still do that in the Semrush app.

Unlike Moz Data or Sitebulb MCP, which expose a long, fixed list of individually named tools, Semrush MCP works through a small, generic set of them. Its own troubleshooting guide names the pattern: discovery tools such as domain_overview, organic_research, keyword_research, competitors_research, backlinks_research, audience_research, traffic_overview, paid_search_research, shopping_research, position_tracking, site_audit and projects, plus a get_report_schema tool that returns the parameters for a given report and an execute_report tool that runs it. In effect, the assistant asks “what can this report take?” before it asks the report itself, which is a sensible way to cover a data surface as wide as Semrush’s without hand-writing dozens of tool definitions.

If MCP itself is new to you, our introduction to MCP for marketers covers the basics. The one-line version: MCP is the open standard that lets an AI assistant call outside tools and pull their data into a conversation, so the assistant answers from Semrush’s live index rather than its general knowledge or a stale export.

What it can’t do

• It cannot write to your Semrush account. No tool creates, edits or deletes a project, a tracked keyword, a report or anything else. The Projects API is exposed strictly read-only.
• It cannot bypass your subscription. Access to SEO data needs an eligible Standard API plan; access to traffic and market data needs a separate Trends API subscription. An assistant connected to Semrush MCP can only see what your plan already entitles you to see.
• It does not include the Local API. Semrush’s Listing Management and Google Business Profile tools, which do include write actions, are not part of Semrush MCP’s scope.
• It has no free tier. Unlike some of the other SEO MCP servers we have covered, there is no throttled free-access mode to try before you subscribe; Semrush’s FAQ is direct that using the MCP “requires an eligible Semrush subscription”.
• It does not cache or reduce cost. Every tool call spends the same API units as calling the endpoint directly. An assistant that chains several report calls to answer one broad question will spend units accordingly.

Getting access: plans and pricing

MCP access is bundled into existing Semrush subscriptions rather than sold separately, which is a simpler model than Moz’s row-based API pricing but means the entry cost is higher; there is no low-cost starter tier built for MCP specifically.

Every MCP tool call spends the same API units as calling the underlying API directly, and Semrush is explicit that “there are no additional MCP usage fees” beyond that. Before letting an agent loose on it, it is worth reading how many units a report you use often actually costs, the same discipline we recommended for Moz’s row-based pricing in our Moz Data MCP review.

Setting it up

Every client uses the same MCP endpoint, https://mcp.semrush.com/v2/mcp, over streamable HTTP, and defaults to OAuth. Clients without OAuth support take a Semrush API key instead, sent as Authorization: Apikey YOUR_API_KEY, found on your account’s API dashboard.

Claude and Claude Code

In Claude, go to Customize → Connect your apps, search for Semrush, click Connect, review the permissions and click Approve. Tag it as @Semrush in a chat to use it.

In Claude Code, add it as a remote server and authenticate from the /mcp menu:

claude mcp add semrush https://mcp.semrush.com/v2/mcp -t http

Then run /mcp, select Semrush MCP, and select Authenticate to complete the OAuth sign-in.

ChatGPT and Perplexity

Semrush ships as an official app in both. In ChatGPT, go to Account → Settings → Apps, find Semrush, click Connect, then Approve on the permission screen; use it in a chat by tagging @Semrush. In Perplexity (paid accounts only), go to Customize → Connectors, find Semrush, click Add connector, and approve the access request.

Cursor and VS Code

Both take the same URL in a small JSON block. In Cursor, under Settings → Cursor Settings → Tools & MCP:

{
  "mcpServers": {
    "semrush": {
      "url": "https://mcp.semrush.com/v2/mcp"
    }
  }
}

In VS Code, create .vscode/mcp.json in your workspace:

{
  "servers": {
    "semrush": {
      "url": "https://mcp.semrush.com/v2/mcp",
      "type": "http"
    }
  }
}

Both trigger the OAuth redirect the first time a tool is called.

Antigravity and Lovable

Google is retiring Gemini CLI in favour of Antigravity; Semrush MCP follows that move and documents Antigravity 2.0 and Antigravity CLI instead, both reading a shared mcp_config.json and authenticating with OAuth, no API key required. Lovable is different again: basic Semrush data (keyword ideas, domain analysis, competitive insights) is already built into Lovable’s chat with no Semrush account needed, and connecting a free or paid Semrush account through Lovable’s Connectors unlocks the full Projects, backlinks and position-tracking data.

Growth Method

Because this is a standard remote MCP server with OAuth, it also works in clients Semrush does not name explicitly. Connect it to Growth Method as a custom integration using Sign in (OAuth) and the /v2/mcp URL, the same way we connected Moz Data and Sitebulb. It sits alongside Sitebulb and DataForSEO in the technical SEO and search data group, so an agent planning a campaign can pull a keyword gap or a competitor’s traffic estimate the same way it pulls sessions from PostHog or impressions from Search Console.

Semrush MCP vs the community servers and DataForSEO MCP

Semrush is not the only way to get this kind of data into an AI client. Two community-built servers predate and now sit alongside Semrush’s own, and DataForSEO offers an official server covering an overlapping but different SEO data set.

The practical difference is control versus convenience. The official server is the obvious default for most teams: no hosting, sign in with the account you already pay for, and it stays current as Semrush ships new reports. The community servers exist mainly for people who want a local CLI alongside the MCP tools (mrkooblu/semrush-mcp bundles a semrush command-line tool with its server) or who are integrating Semrush into a coding agent’s shell rather than a chat client. DataForSEO MCP is a genuinely different data source, not a Semrush substitute, and worth pairing rather than choosing between if your team already holds both subscriptions; see our Sitebulb MCP review for how a technical-crawl data source and an SEO-metrics data source complement each other in the same assistant.

Prompts worth stealing

Semrush’s own MCP page and use-case guide publish full prompts rather than a tool reference, which fits a server built around report execution rather than dozens of narrow tools:

• Opportunity scan: “Using Semrush data for [country], analyze [niche]: build a keyword strategy, identify top organic competitors and content gaps; then summarize and prioritize.”
• Content refresh: “Improve existing content and find new opportunities for [domain] based on search trends and high-impact keywords: identify the 10 most critical pages to update, add or remove, suggest improvements for each, and propose new topics from the top 20 missing keywords.”
• Reporting: “Using Semrush [country] data, generate a monthly SEO report for [domain]. Include estimated organic traffic, top landing pages, the biggest keyword gains and losses, and 3 recommended actions.”
• Competitive check: “Who’s winning traffic among [category] in [country] this month?”
• Cross-tool: “Take our top 10 pages by organic sessions from PostHog and fetch their Semrush ranking keyword count and Domain Authority equivalent.”

That last style of prompt is where this earns its place for a marketing team rather than a single SEO specialist. Semrush MCP only ever returns Semrush data, but your assistant can reason across every tool you have connected to it. Pair it with Search Console or the GA4 MCP and “which of our pages rank well but convert poorly” becomes a single question instead of two exports and a spreadsheet join.

Security: what a reviewer will want to know

Semrush has not published a dedicated security overview for the MCP the way Sitebulb has, so the following comes from the developer documentation and the shape of the server itself:

• OAuth is the default. Sign-in happens on Semrush’s own login page, not inside the AI client, and the client never sees your Semrush password.
• API keys are the fallback, not the default. Only clients without OAuth support use the header-based key, and Semrush’s own account dashboard is where you generate and rotate it.
• Read-only by scope. The SEO and Trends APIs only ever return data. The Projects API is deliberately restricted to its read-only methods through MCP, even though the underlying API can create and update projects.
• Unit quota is the blast radius. A connected assistant can read anything your subscription entitles it to and spend your API units doing it. It cannot delete or modify a project, a report or an account setting.
• What happens after the data lands in your chat is governed by Anthropic, OpenAI, Google or whichever client you use, under your agreement with them, not by Semrush. Semrush’s own disclaimer on the MCP documentation page is blunt about the other side of that: AI-generated responses “may be incomplete or inaccurate”, so verify before publishing or acting on them.

Getting started checklist

1. Confirm your Semrush plan includes MCP access: Semrush One Starter or above, or SEO Classic Pro or above, for SEO data; a Trends API subscription for traffic and market data.
2. Pick your client and connect: the app marketplace inside Claude, ChatGPT or Perplexity, the URL directly in Cursor or VS Code, or the claude mcp add command in Claude Code.
3. Complete the OAuth sign-in, or generate an API key from your Semrush account if your client needs one.
4. Run a one-unit warm-up prompt: “What’s the Domain Authority equivalent and organic traffic estimate for growthmethod.com?”
5. Ask the opportunity-scan prompt for your own site and turn the best answer into a campaign.
6. Connect the tools that will measure the result (Search Console, GA4 or PostHog) to the same assistant, so Semrush finds the opportunity and your analytics proves whether the change worked.

About Growth Method

Growth Method is the agentic marketing platform for B2B teams. Pre-built AI agents plan, launch and analyse marketing campaigns from your live marketing data, with your martech stack connected in minutes. Semrush is already one of the platforms on our own homepage integrations grid, alongside PostHog, Google Ads and Webflow: connect it once, and a Semrush keyword gap or competitor traffic shift becomes a prioritised campaign, launched with its baseline recorded, and measured on organic sessions rather than on a metric that sits inside Semrush alone.

We are on-track to deliver a 43% increase in inbound leads this year. There is no doubt the adoption of Growth Method is the primary driver behind these results.

Laura Perrott, Colt Technology Services

If you would rather run more campaigns than manage exports, get started.

Further reading

Semrush’s own documentation, in the order you will probably need it:

• Semrush MCP overview
• Semrush MCP developer documentation
• Getting started with Semrush MCP
• What is Semrush MCP?
• 16 Semrush MCP prompts for Claude or ChatGPT
• How Semrush MCP connects your AI tools to live marketing data

Related reading from us:

• Moz Data MCP server: Domain Authority and keyword data in Claude and ChatGPT
• Sitebulb MCP server: chat with your crawl data in Claude and ChatGPT
• MCP vs API: what marketers need to know
• An introduction to MCP for marketers
• GA4 MCP server: connect Google Analytics to Claude and Cursor

Frequently asked questions

What is the Semrush MCP server?

Semrush MCP is Semrush’s official, hosted Model Context Protocol server. It runs at https://mcp.semrush.com/v2/mcp over streamable HTTP and gives AI assistants secure access to Semrush’s SEO API, Trends API and the read-only parts of its Projects API, so you can ask for domain overviews, keyword research, backlink data, traffic estimates and more in plain English instead of exporting reports.

Is the Semrush MCP server official?

Yes. It is built and hosted by Semrush, listed as an official app inside Claude, ChatGPT and Perplexity, and documented on developer.semrush.com. That is distinct from community-built servers such as metehan777/semrush-mcp and mrkooblu/semrush-mcp, which are third-party projects that call the Semrush API themselves and require you to hold and manage your own API key.

How does the Semrush MCP server authenticate?

Two ways. OAuth is the default: when you add the server or approve the app inside Claude, ChatGPT, Cursor or another client, you are redirected to a Semrush login page to sign in and approve access, and no key is stored in the client. If your client does not support OAuth, you can instead pass a Semrush API key as a Bearer-style token in the Authorization header, in the format Authorization: Apikey YOUR_API_KEY.

Which Semrush plan do I need for MCP access?

For SEO data, you need Semrush One Starter, Semrush One Pro+, SEO Classic Pro or SEO Classic Guru, all of which include 50,000 API units that refresh on your renewal date, or SEO Classic Business or Semrush One Advanced with an added API units package. For Trends data (traffic and market estimates), any Trends API subscription, Basic or Premium, includes MCP access. Semrush MCP requires an eligible subscription; there is no free tier.

Is the Semrush MCP server read-only?

Yes, for data. The SEO and Trends APIs only ever return data, and the Projects API is exposed read-only through MCP, meaning you can pull data from your existing Semrush projects but cannot create, modify or delete a project through the MCP connection. Every tool call spends the same API units as a direct API call, so there is nothing to break, but there is a cost to run through.

Which AI clients work with Semrush MCP?

Semrush documents Claude and Claude Code, ChatGPT, Cursor, VS Code, Perplexity (paid accounts only), Antigravity 2.0 and Antigravity CLI, and Lovable. Gemini CLI is being retired in favour of Antigravity. Because it is a standard remote MCP server with OAuth, it also works in other clients that support remote connectors, including Growth Method’s custom integrations.

How does the official Semrush MCP compare to the community Semrush MCP servers?

The official server is hosted by Semrush, needs no installation, authenticates with OAuth or your existing API key, and covers the full SEO API, Trends API and read-only Projects API behind a small set of discovery tools. Community servers such as metehan777/semrush-mcp and mrkooblu/semrush-mcp are self-hosted Node projects you install and run yourself, authenticate with an environment-variable API key only, and typically expose a fixed, narrower set of tools such as domain overview, keyword overview and backlinks.