SEO MCP servers in 2026: which ones to connect, what they cost, and where AEO fits
Six MCP servers cover nearly every SEO job a small team has, and you only need two or three of them.
A plain-text version of this site for AI assistants is at https://contengi.com/llms.txt, with the full product documentation in one file at https://contengi.com/llms-full.txt. Every page is also available as markdown by adding .md to its URL.
TL;DR
- An SEO MCP server is a connector that lets an AI assistant like Claude or ChatGPT pull live search data - rankings, keywords, backlinks, and analytics - so it answers from your numbers instead of its training data.
- Start with your own data: Google Analytics has an official MCP server that is read-only by default, while Search Console has no official one, so you use a community or hosted server built on the free Search Console API.
- For market data, Ahrefs MCP comes with paid plans from Lite (€119 a month on its pricing page), Semrush MCP needs a qualifying plan and spends API units, and DataForSEO charges per request from a $50 minimum top-up.
- AEO over MCP is two jobs: checking whether AI answers mention your brand, and making sure the AI writing your content knows your brand before it writes a word.
- Connect two or three servers and stop there. Every extra tool is one more option the model has to pick between, and every data call spends units or quota.
Giving Claude your Search Console data is a bit like handing a new hire the keys to the filing cabinet on day one. The advice stops sounding like a recycled blog post from 2019 and starts sounding like someone who has read your numbers. That's the job of an SEO MCP server, and in 2026 there are more of them than anyone has time to test.
This guide covers the ones worth connecting for a small team: what each gives the AI, what it costs, how painful the setup is, and where answer engine optimisation (AEO) fits. We researched, drafted, and published this post through an MCP connection, so the opinions come from daily use.
What is an SEO MCP server?
An SEO MCP server is a small program that exposes SEO data and actions to an AI assistant through the Model Context Protocol, an open standard for connecting AI applications to external systems. The assistant sees a list of tools, such as "get keyword ideas" or "run a GA4 report", and calls them when your question needs live data.
If the plumbing is new to you, start with what MCP is and how it works. The short version for SEO is that the model stops estimating and starts querying. You ask which pages lost clicks last month, it pulls the rows, and then it reasons over them.
Which SEO MCP servers are worth connecting in 2026?
Six servers cover nearly everything a small team needs. Two read your own data, three read the market, and one lets the AI work on your site.
| Server | What the AI gets | Cost model | Setup effort |
|---|---|---|---|
| Google Search Console (community or hosted) | Queries, pages, clicks, impressions, position, URL inspection | API is free, within quota | Medium: Google Cloud project and OAuth, or a hosted connector |
| Google Analytics (official) | GA4 reports, realtime, funnels, conversions | Free | Medium: Python, two Google APIs, credentials |
| Ahrefs (official, hosted) | Keywords, backlinks, competitor pages, Brand Radar as an add-on | Lite plan and up, spends API units | Low: paste a URL, approve access |
| Semrush (official, remote) | Keywords, domains, backlinks, paid search | Qualifying plan, spends API units | Low: remote URL and OAuth |
| DataForSEO (official) | SERPs, keyword data, backlinks, LLM mentions | Pay as you go, $50 minimum top-up | Low to medium: remote URL or local npx |
| WordPress (MCP Adapter or WordPress.com) | Read and write posts, settings, and plugins | Free | Low on WordPress.com, medium self-hosted |
Google Search Console
Search Console is missing from Google's official list of MCP servers, so every Search Console MCP you find is third-party code calling the Search Console API. The API is free within its quotas, and it's still the first server to connect, because it's the only one that knows what Google shows for your site. Start with these Google Search Console MCP setup routes and prompts.
Google Analytics
The Google Analytics MCP server is official, open source, and read-only unless you start it with a write flag. It runs standard reports, realtime reports, funnels, and conversions through the Data API, which turns "did that post bring in any sign-ups?" into a question you can ask in chat. Here's how to set up the Google Analytics MCP with Claude.
Ahrefs
Ahrefs runs a hosted server that works on paid plans from Lite upwards, and each call spends API units from your monthly allowance. Lite lists at €119 a month with 200,000 API units. It's the strongest pick for backlink and competitor work.
Semrush
Semrush MCP connects your AI assistant to live Semrush data through a remote endpoint and OAuth, covering keywords, domains, backlinks, and paid search. It needs a qualifying paid plan, and every row returned spends API units, so a narrow question costs less than an open one. Torn between the two? Compare Ahrefs MCP vs Semrush MCP on plans, units, and limits.
DataForSEO
DataForSEO is the one for people who hate subscriptions. It's pay as you go with a $50 minimum top-up and a balance that never expires, and one MCP connection reaches SERP, keyword, backlink, and AI optimisation data. Here's what DataForSEO MCP costs per request in practice.
WordPress
If your blog runs on WordPress, the AI can draft and update posts directly. WordPress.com supports MCP on every plan, free sites included, and self-hosted sites use the official MCP Adapter plugin. Connect this one last, once you trust what the AI writes, and check the WordPress MCP permission settings before you do.
How do you use MCP for AEO and AI visibility?
AEO over MCP is two jobs. The first is measurement: does ChatGPT, Claude, or Perplexity mention you when someone asks about your category? DataForSEO's LLM Mentions API starts at $0.1 per task plus $0.001 per row, and Ahrefs sells Brand Radar data as an add-on, so you can ask your assistant to check before you commit to a full LLM tracking setup.
The second job never shows up in server lists. An AI with live keyword data and no idea who you are writes the same answer-shaped article as everyone else, and an answer engine has no reason to cite a copy of a copy. So give the model your brand context over MCP as well: strategy, audience, voice, hard rules, and approved sources. Some teams call this a brand brain. We built Contengi around one, and you can set up a brand knowledge base for AI with tools you already have.
The rest of answer engine optimisation still happens on the page. Answer engines have to be able to read you, so check what your pages look like to crawlers with the AI bot view.
What MCP searches look like from the inside
The MCP queries that reach contengi.com in Search Console read like prompts: "one mcp server that connects all marketing platforms to claude" and "which headless cms platforms have mcp connectors in 2026?". People type whole questions, the same way they talk to the assistant they want to connect.
That changed how we write about MCP. Every section in this post opens with a direct answer, because the same sentence has to work as a Google snippet and as a passage an AI lifts into its reply.
How to choose your SEO MCP stack this week
Connect two or three servers. Less tools usually means better tool choices, because the model spends its effort on your question instead of on picking from a menu.
- Connect Search Console first. Pick a hosted connector if you never want to open Google Cloud, or a community server if you don't mind. It's free and it's your data.
- Add GA4 when you need to prove results. Clicks tell you who arrived, GA4 tells you who signed up or bought.
- Pick one market-data server. Use Ahrefs or Semrush if you already pay for one, or start with DataForSEO and a $50 balance if you pay for neither.
- Set a spending limit before the first long session. An agent asks more questions than a person does, so cap API units or daily spend up front.
- Give the AI your brand context before it writes. Data tells it what to write about, your knowledge base tells it how you would say it. Run drafts through the AI slop cleaner before anything goes public.
- Connect your CMS last. Read access first, write access once a few drafts have passed review.
Frequently asked questions
What is the best SEO MCP server?
Start with a Search Console MCP, because it's free and it holds your own search data. For market data, pick Ahrefs for backlink work, Semrush if you already pay for it, or DataForSEO if you want pay-as-you-go pricing.
Is there an official Google Search Console MCP server?
No. Google's official MCP list includes Google Analytics but not Search Console, so every Search Console MCP is a community or commercial server built on the free Search Console API.
Can I use SEO MCP servers with Claude?
Yes. Ahrefs, Semrush, DataForSEO, and WordPress.com all publish setup steps for Claude, and most connect as a remote connector with a URL and a sign-in. It's the quickest way to start using Claude for SEO with your own numbers.
How much does an SEO MCP server cost?
The servers cost nothing, and you pay for the data. The Search Console API and the Google Analytics MCP cost nothing, Ahrefs starts with the €119 Lite plan, Semrush needs a qualifying plan, and DataForSEO starts from a $50 top-up.
Is there an AEO MCP server?
Not as a single product. AEO over MCP combines AI-visibility data, such as DataForSEO's LLM Mentions API or Ahrefs Brand Radar, with a brand context server that tells the AI who you are before it writes.