A GEO MCP is an MCP server that plugs generative engine optimisation tools into your AI agent, so the agent can audit a page for AI search visibility on its own.
MCP stands for Model Context Protocol, which is an open standard for connecting AI apps like Claude Code or Hermes Agent to outside tools.
GEO stands for generative engine optimisation, which is the work of getting your business named when someone asks Google AI, Gemini or ChatGPT a question.
So a GEO MCP is simply a toolbox of GEO checks that your agent can call, such as crawler access checks, schema checks and citation-readiness scores.
I read the README of every real GEO MCP I could find on GitHub today, and I'll be straight with you: they're all small community projects with between zero and three stars.
My number one pick below is my own Agent OS, so yes, this is my own list, and I'll explain exactly why it sits above the repos.
If you'd rather have my team look at your AI visibility with you first, book a free one-to-one GEO, AEO AI audit and get a custom game plan.
The Best GEO MCP Options Ranked
Here's the ranking table first, so you get the answer in five seconds.
| Rank | Option | What it is | Best for |
|---|---|---|---|
| 1 | Agent OS (my own, inside the AI Profit Boardroom) | My agent dashboard with the Search Console workflow and SEO skill from my video. | Operators who want the full GEO workflow, not just audit tools. |
| 2 | Hermes Agent on its own | The free, open-source agent running the same Search Console process without my dashboard. | People who want the free DIY version of the same workflow. |
| 3 | tmolavi/mcp-geo-server | A Python MCP server that exposes the SAGE audit engine across technical SEO, entity AEO and GEO readiness. | Deep page audits that label every finding with an evidence level. |
| 4 | jpurnell/GeoSEOMCP | A Swift MCP server with 29 GEO analysis tools that score text you supply, fully offline. | Mac users who want repeatable offline scoring. |
| 5 | SEOSiri-Official/aeo-geo-mcp | A local-first Python MCP server with 10 tools for AEO, GEO, llm.txt and schema checks. | Quick llm.txt and schema density checks. |
| 6 | henu-wang/geoscore-mcp | A JavaScript MCP server that wraps the hosted GEOScore scanning service. | People who want a hosted scan, if the service is reachable. |
I rank my own Agent OS first because a GEO MCP only gives your agent a set of checks, while the hard part of GEO is deciding what to publish and then publishing it.
The repos are ranked by how much their READMEs actually cover, how recently they were updated and how clearly they explain their limits.
I haven't run these four repos myself, so the descriptions below come straight from each README and the GitHub data I pulled on 19 October 2026.
🔥 Want my exact GEO MCP Setup setup, step by step? Inside the AI Profit Boardroom you get step-by-step video trainings, my installable Agent OS and four coaching calls a week with 3,400+ members building real automations. → Join the AI Profit Boardroom
What A GEO MCP Actually Does
Think of your agent as a smart worker with no tools in its hands, and an MCP server as the tool belt you hand it.
A GEO MCP is a tool belt built for one job, which is checking how ready a page is to be read, understood and cited by AI search.
Typical GEO MCP tools do things like these.
- They check your robots.txt to see whether AI crawlers such as GPTBot, ClaudeBot and PerplexityBot are allowed in.
- They validate your JSON-LD schema and score how complete it is.
- They score passages for how easy they are to quote, which some READMEs call citability.
- They check or generate an llms.txt file.
- They check heading structure, meta tags and other technical SEO basics.
That's only the audit half of GEO, because none of these tools can tell you which topic to write next or give you a real case study to write about.
Why My Own Workflow Doesn't Use A GEO MCP
In my video above, I show the exact GEO workflow I run, and it doesn't use a GEO MCP at all.
It uses three things, which are an agent, my own Google Search Console data and an SEO skill.
I can run it from my Agent OS, or from Hermes Agent directly, because it's the same process either way.
Hermes Agent is free and open source, and I use OpenCode as the free brain.
When I recorded the video, DeepSeek V4.1 Flash was free inside OpenCode.
Here's the three-step workflow from the video.
- I export my Google Search Console performance data to a CSV file, which is free.
- I give that CSV to Hermes and ask it to analyse the data and, based on what's actually working, give me keyword and topic ideas to rank inside AI search engines.
- I give Hermes my SEO skill plus access to my sites, and it writes and publishes an article from a case-study angle.
In step two, Hermes builds a report in the workspace that shows clicks, brand share of clicks, non-brand click-through rate and the month's clicks.
It also lists what's working, such as comparison articles, Agent OS topics, the free AI community and Hermes topics, and gives a verdict on the best pages.
I'll be honest that this is Google organic data, because Google and the AI engines don't report AI traffic well yet.
The audit tools look at a page, but your Search Console data tells you what Google already trusts you for.
📞 Want my team to read your Search Console data with you? Book a free AEO SEO strategy session and we'll look at your site, your competitors and the keywords you're closest to winning. → Get your free one-to-one GEO, AEO AI audit
Before You Install Any GEO MCP: The Safety Review
An MCP server is code that runs on your machine or on someone else's server, and your agent can call its tools on your behalf.
Anthropic's Claude Code docs say to verify you trust each server before connecting it, and they warn that servers that fetch outside content can expose you to prompt injection.
The Hermes docs say the same kind of thing about their own catalogue, which is that you should still read what an entry installs and runs before you install it.
Here's the review I'd do on any GEO MCP before it goes anywhere near my agent.
- Read the whole README and check that it explains what every tool does.
- Open the source code and look at which tools fetch URLs, write files or call outside services.
- Check the licence, the last update date and whether anyone else uses it.
- Check what the server can reach, and start it with the fewest permissions and tools it needs.
- Test it on a throwaway project first, not on your live sites.
- Keep the agent's ability to publish separate from any MCP you're still testing.
In my plugins video below, I make the same point about Hermes plugins, which is that you should check the GitHub repository and README before you install anything community-made.
If you want the wider plugin picture, my Hermes Agent plugins marketplace walkthrough covers that safety check in more detail.
How To Install A GEO MCP In Claude Code
These steps come from Anthropic's Claude Code MCP docs, and they work the same way for any MCP server.
For a remote server with an HTTPS URL, the docs use this command.
claude mcp add --transport http <name> <url>
For a local server that runs on your machine, the docs put the launch command after a double dash.
claude mcp add <name> -- <command> [args...]
By default the server is saved to local scope, which means it only loads in the current project.
The docs say you can add --scope project to share it through a .mcp.json file, or --scope user to load it in all your projects.
Then you can check the connection with these commands.
claude mcp list
claude mcp get <name>
Inside Claude Code you can also type /mcp to see your servers and their status.
If a GEO MCP README only gives you a config block for Claude Desktop or Cursor, the Claude Code docs explain how to pass the inner part of that block to claude mcp add-json.
How To Install A GEO MCP In Hermes Agent
These steps come from the Hermes Agent MCP docs.
The docs say MCP support ships with the standard install, so there's no extra step to turn it on.
You add a server under mcp_servers in your ~/.hermes/config.yaml file.
A local server takes a command and args, which come from the server's own README.
mcp_servers:
geo:
command: "<command from the README>"
args: ["<args from the README>"]
A remote server takes a url instead.
Then you start Hermes with hermes chat and ask it to use the new tools.
The docs say Hermes discovers the server's tools at startup and uses them like any other tool.
There are three extra things in the Hermes docs that I'd use with a GEO MCP.
- You can whitelist tools with
tools: include:under the server, so only the tools you've reviewed get registered. - You can run
hermes mcp test <name>to see what the server actually answered if it won't connect. - You can use
/reload-mcpto pick up a change without starting a whole new session.
If you already set the server up in Claude Code, the docs say hermes import-agent claude-code migrates your MCP servers across.
If you want to see MCP working the other way round, my Hermes MCP server guide shows how Codex connects to Hermes.
🚀 Ready to put GEO MCP Setup to work in your business? Inside the AI Profit Boardroom you get step-by-step video trainings, my installable Agent OS and four coaching calls a week with 3,400+ members building real automations. → Join the AI Profit Boardroom
The Four GEO MCP Repos, From Their READMEs
Here's what each repo says about itself, with the GitHub numbers I pulled on 19 October 2026.
| Repo | Language | Stars | Licence | Last code push | Tools |
|---|---|---|---|---|---|
| tmolavi/mcp-geo-server | Python | 3 | MIT | 18 September 2026 | Full audits, three pillar tools, entity and citation tools, llms.txt generator. |
| jpurnell/GeoSEOMCP | Swift | 0 | MIT in the README | 17 October 2026 | 29 tools across nine groups. |
| SEOSiri-Official/aeo-geo-mcp | Python | 0 | MIT | 2 October 2026 | 10 tools. |
| henu-wang/geoscore-mcp | JavaScript | 1 | MIT in the README | 5 March 2026 | 8 tools, some needing an API key. |
Here's the one setup note per repo that an operator needs, taken from each README.
- The tmolavi server audits a live URL or raw HTML across technical SEO, entity AEO and GEO readiness, and its README installs it with pip or uv, but I couldn't find the package on the public PyPI index on 19 October 2026, so you may need the repo source.
- GeoSEOMCP needs Swift 6.2 and macOS 14 or newer, installs with
git clone,swift buildandswift run geoseo-mcp-server, and never fetches a URL itself, so your agent passes the page text in. - The SEOSiri server installs from the repo with
pip install -e ., and its README gives a Claude Desktop config block you can convert withclaude mcp add-json. - The GEOScore server needs an API key for its deep scan, fix generator and citation check, its code hasn't been pushed since 5 March 2026, geoscoreai.com returned an error when I tried it on 19 October 2026, and I couldn't find the package on npm that day.
A Note On llms.txt Before You Trust Any Score
All four of these repos check or generate an llms.txt-style file, although the SEOSiri README calls it llm.txt.
Google's AI features and your website page says you don't need AI text files or special schema to appear in AI Overviews or AI Mode, so I treat any llms.txt score as optional.
📞 Not sure which checks matter for your site? Grab a free strategy session and we'll show you why competitors appear in AI answers more often than you. → Book the free one-to-one GEO, AEO AI audit
The Operator Setup I'd Actually Run
If I were adding a GEO MCP to my own stack tomorrow, this is the order I'd do it in.
- Export your Google Search Console data to CSV and run the analysis prompt from my video first, so you know what's working before you audit anything.
- Pick the five pages that matter most from that report.
- Review one GEO MCP using the safety checklist above, and install it in a test project only.
- Whitelist just the read-only audit tools you need.
- Ask the agent to audit those five pages and list the top three fixes for each.
- Fix the technical issues by hand or with the agent, then re-run the audit.
- Write new pages with your SEO skill from a real case-study angle, because that's what gets you named.
- Fact-check every page by hand before it goes live.
Step eight isn't optional.
Google's page on using generative AI content was updated on 1 October 2026, and it says it's "critical to manually factcheck and review all AI-generated content" before publishing.
That review also covers titles, meta descriptions, structured data and image alt text.
What This Workflow Has Done For Me
Here's what I show on screen in my video, and these are my own results, not a promise.
- For "best geo expert", my content shows in the Google AI answer at the top and ranks first organically.
- In AI Mode, asking for the best GEO experts to follow in 2026 lists me first and links to my sites.
- One of my sites went from zero to more than 1,300 clicks a day, and two other sites are on the same path.
A GEO MCP can help you check your pages, but it can't run your tests for you, and real tests are what get you named.
Where The AI Profit Boardroom Fits
🔥 Want my Agent OS, SEO skill and the full GEO workflow? Inside the AI Profit Boardroom, you get the Agent OS, my SEO skill checklist, the AI SEO automation section and Hermes trainings that include Hermes with Claude MCP, plus four coaching calls a week with 3,400+ members. → Get access here
My SEO skill is a checklist of roughly 100 points, and it writes everything from a case-study angle.
The Agent OS runs a crew of AI agents under one shared memory, and it isn't on GitHub because it's only available inside the Boardroom.
Limits Of Any GEO MCP
These are the honest limits I'd keep in mind before you rely on one.
- Every GEO MCP I found is a small community project, so support and updates aren't guaranteed.
- A GEO score is a proxy, and no tool can tell you what an AI answer will say tomorrow.
- Nobody can guarantee you'll be named in an AI answer, and that includes me.
Related Reading
- Run a free GEO audit before you install anything.
- See four real GEO examples from my own screens.
- Follow the step-by-step guide to doing GEO for your website.
- Learn how I use Claude for GEO and SEO.
Also On Our Network
- 🌐 How agencies and businesses put a GEO MCP to work
- 🌐 Every GEO MCP server rated feature by feature
- 🌐 A GEO MCP quick start for your Agent OS
- 🌐 Which AI models are best at driving a GEO MCP
FAQ: GEO MCP
What is a GEO MCP?
A GEO MCP is a Model Context Protocol server that gives an AI agent generative engine optimisation tools.
Those tools include AI crawler checks, schema checks and citation-readiness scores, so the agent can audit pages for AI search visibility.
Which GEO MCP servers exist right now?
The real ones I found on GitHub on 19 October 2026 are jpurnell/GeoSEOMCP, tmolavi/mcp-geo-server, SEOSiri-Official/aeo-geo-mcp and henu-wang/geoscore-mcp.
All four are small community projects with between zero and three stars.
How do I install a GEO MCP in Claude Code?
Anthropic's docs use claude mcp add with a name and either a URL for a remote server or a launch command after a double dash for a local server.
You then check it with claude mcp list or the /mcp command inside Claude Code.
How do I install a GEO MCP in Hermes Agent?
The Hermes docs say you add the server under mcp_servers in ~/.hermes/config.yaml with a command and args, or a url for a remote server.
Then you start Hermes and it discovers the tools at startup.
Do I need a GEO MCP to do GEO?
No, and my own workflow in my video uses an agent, a Search Console CSV export and an SEO skill with no GEO MCP at all.
An MCP is just one way to give your agent extra tools.
Is it safe to install a GEO MCP server?
It's only safe after you've reviewed it, because an MCP server runs code and its tools act on your behalf.
Read the repository, check what each tool can reach, and only connect servers you trust.
📺 Video notes + links to the tools 👉
🎥 Learn how I make these videos 👉
🆓 Get a FREE AI Course + Community + 1,000 AI Agents 👉
About Julian
I'm Julian Goldie, an SEO entrepreneur, author and founder of the AI Profit Boardroom, which has 3,400+ members.
I help business owners scale with AI agents, automation and SEO.
- I started Goldie Agency on my own, and it is now a 7-figure SEO and link-building agency with 70+ people.
- I've grown my YouTube channel to 400,000+ subscribers.
- I wrote "Link Building Mastery".
- My Udemy courses have taught 29,000+ students.
→ Get my best AI training inside the AI Profit Boardroom
Start with your Search Console data, review any server before you trust it, book a free strategy session if you want a second pair of eyes, and then add a GEO MCP only where it saves you real work.
