Domain and traffic overviews
domain_overview and traffic_overview are the two entry points a model reaches for first, and the ones that answer the question most prompts actually start with: how big is this site.
Semrush MCP Server
The official Semrush MCP server: one HTTPS endpoint, two ways to authenticate, and a tool list that maps onto Semrush's own reports. Endpoint, commands, tools and plan requirements read from Semrush's developer docs and knowledge base on 2 September 2026.
Interactive examples · No account is connected on this page
Updated
Semrush MCP is Semrush's own Model Context Protocol server, at https://mcp.semrush.com/v2/mcp, over streamable HTTP transport only. It authenticates with OAuth by default, or with an API key sent as an Authorization: Apikey header for clients that cannot do OAuth. Calls spend the same API units as the Semrush API, and the reports cover domains, keywords, backlinks, traffic and site audits.
Try a Semrush MCP task
Pick an example to see which Semrush MCP tools a task would call and what the result looks like. Examples are illustrative; nothing on this page connects to Semrush.
Connect Semrush MCP
Semrush MCP is a remote server at https://mcp.semrush.com/v2/mcp over streamable HTTP only. Check first that your subscription includes API units for the data you need. Then add the endpoint to Claude Code, Cursor or VS Code, or connect the Semrush app inside Claude, ChatGPT or Perplexity. Clients without OAuth use an Authorization: Apikey header.
Check that your subscription covers the data you want
This is the step that decides whether anything works. Semrush's docs tie MCP access to API units by data type: SEO data needs Semrush One Starter or Pro and above, or SEO Classic Pro or Guru, with 50,000 units included. Traffic data needs a Trends API Basic or Premium subscription. Projects API access is included, and is read-only.
Add the endpoint to Claude Code
One command registers the remote server. Then open Claude Code, run the /mcp command, select Semrush MCP and select Authenticate to complete the OAuth flow.
claude mcp add semrush https://mcp.semrush.com/v2/mcp -t httpOr connect through the app in Claude, ChatGPT or Perplexity
For the chat products, Semrush ships an app rather than a raw endpoint. Open the Semrush app inside Claude or ChatGPT, click Connect, review the access it asks for and approve it; afterwards you address it as @Semrush in your prompts. Perplexity uses Add connector instead, and the knowledge base notes it needs a paid Perplexity account.
Paste the JSON in an editor
Cursor and VS Code take the endpoint as configuration. The two formats differ in shape: Cursor nests it under mcpServers, VS Code under servers with an explicit type.
{
"mcpServers": {
"semrush": {
"url": "https://mcp.semrush.com/v2/mcp"
}
}
}
{
"servers": {
"semrush": {
"url": "https://mcp.semrush.com/v2/mcp",
"type": "http"
}
}
}Use an API key where OAuth is not available
If your agent does not support OAuth, authenticate with an API key passed in the Authorization header. The scheme word is Apikey, not Bearer — that detail is where most failed connections come from. Semrush's own example is the Gemini CLI.
Authorization: Apikey YOUR_API_KEY
gemini mcp add --transport http \
--header "Authorization: Apikey YOUR_API_KEY" \
semrush https://mcp.semrush.com/v2/mcpOfficial documentation: Semrush MCP developer docs.
The two documented ways to authenticate
Semrush documents two ways to authenticate against the same endpoint and transport. OAuth is the default and needs no extra configuration: the client registers and redirects you to sign in. An API key is for clients that cannot do OAuth and travels in an Authorization header whose scheme word is Apikey, not Bearer, which is where most failed connections come from.
| What differs | OAuth (default) | API key |
|---|---|---|
| Configuration needed | None. The docs state no additional configuration or headers are required. | An Authorization header on every request |
| What happens on connect | The AI agent automatically registers with the OAuth system and redirects you to the Semrush login page | No redirect; the header carries the credential |
| Exact form | claude mcp add semrush https://mcp.semrush.com/v2/mcp -t http | Authorization: Apikey YOUR_API_KEY |
| When to pick it | Clients that support OAuth — the docs cover Claude, ChatGPT, VS Code, Cursor, Antigravity and Perplexity | Clients that do not; the documented example is the Gemini CLI |
| Endpoint and transport | https://mcp.semrush.com/v2/mcp, streamable HTTP only | The same endpoint and the same transport |
Read from developer.semrush.com and semrush.com/kb on 2026-09-02. Both paths consume standard API units per request, priced by report type.
Semrush MCP exposes reports rather than raw endpoints: domain and traffic overviews, keyword, organic and competitor research, paid and shopping research, backlinks and audience data, and read-only access to your own projects. Two utility tools discover and run any other report. Every call spends the same API units as the Semrush API.
Keywords and organic search
keyword_research, organic_research and competitors_research cover the classic Semrush workflow — what a site ranks for, what the terms look like, and who else is competing for them.
Paid and shopping
paid_search_research and shopping_research extend the same idea to ads and product listings, so an agent comparing organic against paid does not have to leave the server.
Backlinks and audience
backlinks_research and audience_research are the off-site half: who links to a domain, and who visits it.
Your own projects, read-only
position_tracking, site_audit and projects reach into Semrush projects you already have. The docs are explicit that this is read-only: MCP cannot create projects or modify settings.
An escape hatch for everything else
get_report_schema and execute_report are the two utility tools. They let a model discover a report's parameters and then run it, which covers reports without a dedicated tool of their own.
Units, not requests, are the budget
Every call consumes standard API units based on report type. An agent left to explore freely will spend them the way any loop spends a metered resource, so cap the exploration before you point one at a large site.
The usual caveat, from Semrush
Semrush's own documentation notes that AI responses may be incomplete or inaccurate and require verification. Live data through MCP does not make the summary on top of it correct.
What is the Semrush MCP server?
Semrush MCP is Semrush’s official Model Context Protocol server at https://mcp.semrush.com/v2/mcp. It lets an AI client run Semrush reports — domains, keywords, backlinks, traffic and your existing projects — over streamable HTTP.
It authenticates with OAuth by default, or with an Authorization: Apikey header for clients that cannot do OAuth. Every call spends the same API units as the Semrush API, and the projects tools are read-only.
Other MCP servers for research and content work
Semrush answers what search engines see. The guides below cover the other sources a research agent usually needs in the same session: community discussion, a browser for pages without an API, a knowledge base and your automation layer. Each guide lists install steps, authentication and the full tool list the server exposes.
Reddit MCP ServersReddit publishes no MCP server of its own. Three community Reddit MCP servers compared on install command, credentials, tools and whether they can post.Playwright MCP ServerHow to install and configure Playwright MCP, Microsoft's server that drives a real browser through accessibility snapshots. Covers npx, HTTP and Docker setup.
Notion MCP ServerHow to connect Notion MCP: the hosted endpoint at mcp.notion.com/mcp, OAuth setup for Claude Code, Cursor and VS Code, the notion- tools and rate limits.
Claude MCPHow Claude MCP works on three surfaces: claude mcp add in Claude Code, .mcpb extensions in Claude Desktop, and remote custom connectors on claude.ai and Cowork.n8n MCP ServerHow n8n does MCP: the instance-level MCP server at /mcp-server/http, the MCP Server Trigger and MCP Client Tool nodes, and the community n8n-mcp package.How to connect Semrush MCP
Check your subscription
The data you can reach depends on your Semrush plan and API units.
Add the endpoint
Add https://mcp.semrush.com/v2/mcp to your client and complete OAuth, or use the API key header.
Start small
Run one domain_overview before letting an agent explore, because every call spends units.
Semrush MCP use cases
Size up a competitor
Start with domain_overview and traffic_overview, then drill into organic_research and backlinks_research.
Plan content from keyword data
Use keyword_research to shortlist terms and check them against what competitors already rank for.
Report on your own projects
Read position_tracking and site_audit results; MCP cannot create projects or change settings.
Combine with community research
Pair Semrush data with Reddit MCP discussions to see why people search for a topic.
Semrush MCP questions
These answers cover what people check before connecting Semrush MCP: whether it is official, what the endpoint is, whether it costs anything beyond your subscription, whether an agent can change your Semrush projects, what to use when a client cannot do OAuth, and which clients Semrush documents.
Is Semrush MCP official?
Yes. It is documented in Semrush's own developer documentation and knowledge base, and served from mcp.semrush.com. That is a real distinction in this corner of the ecosystem: several data tools have only community-built MCP servers, where the endpoint, the tool names and the support all come from a volunteer. This one does not.
What is the Semrush MCP endpoint?
https://mcp.semrush.com/v2/mcp, and the docs state it supports streamable HTTP transport only. There is no stdio option and no local package to install — every client, whether it is Claude Code, Cursor, VS Code or a CLI, points at that same URL.
Does using it cost anything on top of my subscription?
It spends the API units your plan already includes. Semrush's docs describe standard API unit consumption per request based on report type, and tie eligibility to specific plans: Semrush One Starter or Pro and above, or SEO Classic Pro or Guru, with 50,000 units included, for SEO data; a Trends API subscription for traffic data.
Can an agent change my Semrush projects through MCP?
No. Access to the Projects API through MCP is documented as read-only, and the limitations section says you cannot create projects or modify settings via MCP. position_tracking, site_audit and projects read what is already configured in your account; setting those up stays a job for the Semrush interface.
My client cannot do OAuth. What do I use?
An API key in the Authorization header, in the form Authorization: Apikey YOUR_API_KEY. Note the scheme word: it is Apikey, not Bearer, and a client that assumes bearer tokens will fail authentication without saying why. Semrush's documented example of this path is the Gemini CLI.
Which clients does Semrush document?
The knowledge base walks through ChatGPT, Claude, Claude Code, VS Code, Cursor, the Gemini CLI and Perplexity, and the developer docs add Antigravity. Chat products connect through a Semrush app you approve; editors and CLIs take the endpoint directly. Perplexity is noted as requiring a paid account.
Live SEO data is an input, not a decision
Pulling Semrush numbers into a model is the cheap part. Turning them into a repeatable process that publishes, tracks and revises is where the work actually sits: deciding which reports run on a schedule, which changes an agent may draft on its own, and which pages a person reviews before anything goes live.