seoskills.sh
Catalog/Docs/SEO MCP servers: GSC, Ahrefs, DataForSEO, Semrush

SEO MCP servers for Search Console, Ahrefs, DataForSEO and Semrush

Updated

An SEO MCP server connects your AI agent to a source of SEO data, so it can pull Search Console performance, keyword, SERP or backlink data while it works, instead of you pasting exports into the chat. Ahrefs, DataForSEO and Semrush each run an official server. For Google Search Console, the server covered here is a community project that runs on your own machine.

The four servers at a glance

ServerMade byHow it connectsWhat you need
Google Search Console (mcp-gsc)Community, MIT licenseLocal server, started with uvxA Google Cloud project with the Search Console API enabled, plus OAuth or a service account
Ahrefs MCPAhrefsHosted, Streamable HTTPA paid plan, Lite or higher
DataForSEO MCPDataForSEOHosted, or local with npxDataForSEO API access
Semrush MCPSemrushHosted, streamable HTTP onlyAn eligible Semrush plan

In Claude Code, add a server from your terminal with claude mcp add, check it with claude mcp list, and finish any sign-in from /mcp inside a session. A new server is active in the current project only, unless you add --scope user. Each server's tool names load into every session and use context, so connect only the ones you use.

Google Search Console MCP

Who makes it. mcp-gsc is a community project, MIT licensed. It is not an official Google server.

What it gives your agent. Your properties and site details; queries and pages with clicks, impressions, CTR and position, filtered by country, device, query or page; comparisons between two periods; URL inspection for crawl and index status, up to 10 URLs at a time; and sitemap status and submission. Tools that add or remove sites, or delete sitemaps, stay off unless you set GSC_ALLOW_DESTRUCTIVE to true.

What you need. A Google Cloud project with the Search Console API enabled, plus either an OAuth client ID for a desktop app, saved as a JSON file, or a service account key whose email you add to the property as a user with full access. The recommended install also needs uv, for uvx.

How it connects. As a local server Claude Code starts on your machine:

claude mcp add gsc --env GSC_OAUTH_CLIENT_SECRETS_FILE=/full/path/to/client_secrets.json -- uvx mcp-search-console

On first use, a browser window opens for Google sign-in; after that, the token is saved. With a service account, set GSC_CREDENTIALS_PATH to the key file and GSC_SKIP_OAUTH to true instead. To test it, ask Claude to list your GSC properties. If uvx isn't found, use the full path that which uvx prints.

Ahrefs MCP

Who makes it. Ahrefs, as a hosted remote server. Its older local server on GitHub is unmaintained, and Ahrefs advises against it.

What it gives your agent. Access to the Ahrefs API, so answers can use real Ahrefs data. Ahrefs' MCP docs list the tool categories and have setup guides for several AI tools, Claude Code included.

What you need. A paid Ahrefs plan, Lite or higher. Each plan caps the rows an agent gets per request and comes with monthly API units, which MCP calls spend the same way direct API calls do. Workspace admins can cap each key's monthly units.

How it connects. Over Streamable HTTP at https://api.ahrefs.com/mcp/mcp:

claude mcp add --transport http ahrefs https://api.ahrefs.com/mcp/mcp

Then run /mcp in a session and choose Authenticate. Once you approve Ahrefs' consent screen, the connection shows up in your Ahrefs account with its own API key, tagged with MCP scope. You can also create an MCP key by hand. Calling the endpoint from your own scripts or bridges isn't permitted; for that, Ahrefs points you to its public API.

DataForSEO MCP server

Who makes it. DataForSEO. The current version is the v3 server in dataforseo/mcp-server-typescript. The previous server is deprecated, so skip guides that use it.

What it gives your agent. Four tools: docs_index, docs_list_sections and docs_search browse the DataForSEO API documentation, and api_request makes authenticated API requests. The agent reads the docs, picks an endpoint, such as a live Google organic SERP or a backlinks summary, and calls it.

What you need. A DataForSEO account with API access. The hosted server signs you in with OAuth. A local server that your client starts uses your API login and password instead.

How it connects. Hosted at https://mcp.dataforseo.com/v3/mcp:

claude mcp add --transport http dataforseo https://mcp.dataforseo.com/v3/mcp

Finish the OAuth sign-in from /mcp. To run it locally, have Claude Code start npx dataforseo-mcp-server with DATAFORSEO_LOGIN and DATAFORSEO_PASSWORD as environment variables.

Semrush MCP

Who makes it. Semrush, as a hosted server that uses streamable HTTP only.

What it gives your agent. Data from the Semrush SEO API, the Trends API (depending on your Trends API plan) and the read-only methods of the Projects API v3, so the agent can read your projects but not change them. Discovery tools such as organic_research, keyword_research and backlinks_research find reports. get_report_schema returns a report's parameters, and execute_report runs it.

What you need. For SEO data, Semrush One Starter or Pro+, or SEO Classic Pro or Guru, which include 50,000 API units that refresh on your renewal date. SEO Classic Business and Semrush One Advanced need an added API units package. For traffic and market data, you need a Trends API plan, Basic or Premium. MCP requests spend the same API units as direct API calls, with no extra MCP fee.

How it connects. At https://mcp.semrush.com/v2/mcp, with OAuth by default. Clients without OAuth can send an API key in an Authorization: Apikey header.

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

Open /mcp, select the Semrush server and choose Authenticate. You'll be sent to the Semrush login page.

Skills that use this data

A server gives Claude tools, such as a call that returns your top queries. A skill gives it the procedure: which calls to make, what to compare and what to hand back. SEO Rank Tracking works this way with the Ahrefs MCP server: the skill picks the keywords, segments them and sets baselines, and the server supplies the data. Skills vs MCP servers covers the difference in full.

These catalog skills work with Search Console, Ahrefs, DataForSEO or Semrush data:

SkillDataWhat it does
SEO DataforseoDataForSEO MCP serverSERP analysis, keyword research and backlink profiles from live data
SEO Rank TrackingAhrefs MCPPicks and segments keywords to track, sets baselines and alert thresholds
Google Search ConsoleSearch Console, GSC APIAnalyzes Search Console data and interprets search performance
SEO GoogleGoogle's SEO APIsSearch Console analytics, URL inspection and sitemaps, plus PageSpeed Insights and CrUX data
Ahrefs-GSC Backlink ReconcilerAhrefs, Search Console, optional SemrushMerges backlink data into one referring-domain set, with gained and lost domains
Semrush-GSC Rank BlenderSemrush, Search ConsoleOne rankings table flagging disagreements and demand Semrush isn't tracking
SEO Metrics API OrchestratorAhrefs, Semrush, Moz, DataForSEOAuthority, backlink, volume and difficulty metrics in one pass, within a per-vendor credit budget
GSC Bulk Export ManagerSearch Console bulk export in BigQueryCost-guarded queries for full query inventories, click decay and daily anomalies

Only the first two say they use an MCP server. The others name an API or an export, or don't say, so check a skill's SKILL.md for what it expects before you set anything up. More tool-specific skills are in SEO Tool Integration.

Not an SEO data server: the seoskills.sh MCP endpoint

seoskills.sh runs an MCP endpoint too, for a different job. It searches this skill catalog and returns skills, not search data about your site. Its tools are search_skills, get_skill and list_categories, so Claude can find a skill mid-task.

claude mcp add --transport http seoskills https://seoskills.sh/api/mcp

To install a skill it finds, run the command on that skill's page.

Next