This article is part of our MCP Guide: what the Model Context Protocol is, how to use it safely, and how to choose servers, tools and help. Want a second opinion on your setup? Request a free MCP readiness review.
MCP changes SEO work by letting an AI assistant pull live data from Search Console, GA4, Ahrefs and Semrush in one conversation. The biggest change is fewer CSV exports and VLOOKUPs. The parts that need judgment, such as deciding what to write, what to fix first and what to ignore, stay with you.
MCP (Model Context Protocol) is an open standard that connects AI apps to other software through small programs called MCP servers. For SEO, those servers wrap the same APIs your tools already offer. If you are new to the idea, read What Is MCP? first.
Below we cover which SEO data sources have MCP servers as of October 2026, who publishes them, and how a typical SEO workflow shifts. We checked each server against the vendor's docs on 28 October 2026.
Key Takeaways
- Ahrefs and Semrush both run official hosted MCP servers, and every call uses API units from your plan.
- We found no Google-published Search Console MCP server; the options we found are community-built.
- Google's GA4 MCP server is official but labeled experimental, and it is mostly read-only.
- MCP removes much of the export-and-join work in audits and reporting, but it does not replace crawling, testing or editorial judgment.
- Treat AI-produced SEO numbers as a first draft and spot-check them against the source tool.
Which SEO data sources have MCP servers
| Source | Publisher | Type | Notes from the docs |
|---|---|---|---|
| Ahrefs | Ahrefs | Hosted | Lite plan or higher; uses API units |
| Semrush | Semrush | Hosted | SEO API, Trends API, read-only Projects |
| Google Search Console | Community projects | Usually local | Service-account sign-in is common |
| GA4 | Google (googleanalytics org) | Local | Experimental; read-only except annotations |
| HubSpot (content + AEO) | HubSpot | Hosted | AEO needs Marketing Hub Pro or Enterprise |
Ahrefs
Ahrefs' MCP documentation says it "provides a hosted MCP server that allows users with a Lite plan or higher to connect their Ahrefs account to supported AI tools." The connection URL is https://api.ahrefs.com/mcp/mcp.
Key details from that page:
- Each plan caps how many rows an agent can pull per request.
- Calls consume API units from your monthly allowance, the same as direct API use.
- New keys created through OAuth default to 200,000 API units per month. Workspace admins can change that.
- Using the endpoint from custom scripts or standalone clients is "unsupported and not permitted." It is meant for AI tools, not as a general API.
Ahrefs' older local server on GitHub is archived and marked as no longer maintained. Use the hosted one.
Semrush
Semrush's MCP page lists the endpoint as https://mcp.semrush.com/v2/mcp. It covers all SEO API endpoints, Trends API endpoints for your subscription level, and read-only Projects API methods.
Sign-in uses OAuth 2.1 by default, with an API key header as a fallback. The page lists which Semrush One and SEO Classic plans include access. Each request consumes API units based on the report type and amount of data.
Google Search Console
As of 28 October 2026, we did not find a Search Console MCP server published by Google. The options we found are community projects built on the public Search Console API.
For example, ahonn/mcp-server-gsc documents a search analytics tool with up to 25,000 rows, regex filters on query and page, and a "quick wins" option. It uses a Google service account that you add to the property.
Community servers can be useful. Read the code first, limit the service account's access, and pin a version.
GA4
The Google Analytics MCP server is in Google's googleanalytics GitHub organization. It runs standard, funnel and realtime reports. It is labeled experimental, and its only write tool, for annotations, is off unless you pass --allow-write.
For SEO, GA4 answers the "did organic traffic convert?" half of the question that Search Console can't.
HubSpot content and AEO
If your content lives in HubSpot, its remote MCP server can read content analytics and write landing pages, website pages and blog posts. HubSpot also lists read access to AEO data and write access to AEO prompts and recommendations. Those AEO features need Marketing Hub Professional or Enterprise.
What changes in the SEO workflow
Here is how common tasks shift once MCP servers are connected.
| Task | Before MCP | With MCP |
|---|---|---|
| Monthly report | Export GSC, GA4 and rank data; join in a sheet | Ask one question; check the AI's table |
| Quick-win hunt | Filter GSC for high impressions, low CTR | Ask for pages in positions 5–15 with low CTR |
| Competitor gap | Run a gap report, export, sort | Ask Ahrefs or Semrush through the AI, then refine |
| Content brief | Pull keywords, SERP notes and internal data by hand | AI drafts a brief from live keyword data |
| Traffic drop triage | Compare date ranges across three tools | AI compares and lists likely pages |
The pattern is the same each time. MCP removes data gathering. It does not remove the decision.
What MCP does not change
Some SEO work still needs other tools or a person:
- Crawling and rendering. An MCP connection to an API is not a site crawler. You still need a crawler to find broken links, redirect chains and render issues.
- Testing changes. The AI can suggest a title rewrite. Whether it works is something you measure over weeks.
- Editorial judgment. Deciding which topics fit your brand, what you can credibly say and what your readers need is still human work.
- Accuracy checks. AI summaries can misread a date range or filter. Always spot-check key numbers in the source tool.
A sample week with MCP
A realistic weekly routine for an in-house SEO or a small agency:
- Monday. Ask for last week's clicks, impressions and conversions from organic search, by page group. Compare with the week before.
- Tuesday. Ask for queries where you rank 5–15 with low CTR. Shortlist five pages for title and meta tests.
- Wednesday. Pull competitor keyword gaps from Ahrefs or Semrush for one topic cluster. Draft briefs.
- Thursday. Write or edit content. MCP doesn't help much here beyond fact lookups.
- Friday. Check API unit usage in Ahrefs or Semrush so a busy week doesn't eat the month's allowance.
Where Claude Code fits
You can also use MCP servers inside Claude Code, Anthropic's command-line coding agent, rather than a chat app. Claude Code can combine MCP data with files on your machine, such as a crawl export or your site's source code. That makes tasks like "find every page with a missing meta description and draft one" practical.
We wrote a separate guide on that: Claude Code for SEO Specialists: What to Hand Off. Our Claude Code learning hub covers setup.
Risks to manage
- Client data separation. Agencies should keep each client's Search Console and GA4 credentials in separate profiles.
- API cost. Agents make many calls. Set per-key unit limits in Ahrefs, and watch Semrush usage.
- Write access to live pages. HubSpot's server can edit pages and posts. Keep that off until you trust the workflow, and review every change before publishing.
- Community code. Search Console servers we found are community projects. Review them like any dependency.
For a structured way to vet servers, use our checklist for choosing MCP servers. The wider marketing view is in MCP for Marketing Teams.
Next step
To see which SEO tasks to hand to an AI agent and which to keep, read Claude Code for SEO Specialists. For a side-by-side of servers, see Best MCP Servers for Marketing and SEO Teams.
Frequently asked questions
Does MCP replace SEO tools like Ahrefs or Semrush?
No. MCP is a way to reach those tools from an AI assistant. You still need the subscription, and each MCP call uses API units from your plan. What changes is how you ask for the data, not where it comes from.
Is it safe to connect Search Console to an AI through MCP?
It can be, with care. The Search Console servers we found are community-built, so read the code, use a service account with limited access, and keep each client's credentials separate. Start with a property where a mistake would cost little.
Can MCP do a technical SEO audit?
Partly. It can pull indexing, performance and analytics data and help you spot patterns. It does not crawl or render your site on its own. You still need a crawler for broken links, redirect chains and rendering problems.
How much does using Ahrefs or Semrush through MCP cost?
There's no separate MCP fee in the docs we checked, but calls consume API units. Ahrefs says MCP needs a Lite plan or higher and draws on your monthly unit allowance. Semrush lists eligible plans and says each request uses units based on the report and data size.
Will MCP help with AI search visibility (GEO and AEO)?
Indirectly. MCP gives an AI faster access to your search and content data, which helps you find gaps. HubSpot's server also exposes AEO data on higher Marketing Hub tiers. The work of earning citations in AI answers is still content and authority work.
