The SEO MCP server for coding agents
Connect Claude Code, Cursor, Codex or any MCP client to live keyword, SERP, backlink and AI-visibility data, so what your agent writes starts from what already ranks instead of what sounds right.
Connected in under a minute
Let your agent set itself up, or paste the config for your client. Nothing to install and nothing running on your machine.
Paste this into Claude Code, Cursor, Codex or any agent that can read a URL. It reads our skill and writes its own config in the right format.
Read https://omnifamous.com/mcp/skill.md and set up the omnifamous SEO MCP server for this project, then use it to audit my site's search and AI visibility.
Read the skill
Command line (recommended)
- Run the command below in your terminal
- Run /mcp inside Claude Code to confirm the server is connected
claude mcp add --transport http omnifamous https://omnifamous.com/api/mcp \ --header "Authorization: Bearer sk_your_secret_key"
Project config
- Create .mcp.json in your project root
- Paste the configuration below and commit it to share with your team
- Restart Claude Code and approve the server when prompted
.mcp.json
{
"mcpServers": {
"omnifamous": {
"type": "http",
"url": "https://omnifamous.com/api/mcp",
"headers": {
"Authorization": "Bearer sk_your_secret_key"
}
}
}
}Add --scope user to the command to make the server available in every project on your machine.
What is an SEO MCP server?
An SEO MCP server is a Model Context Protocol server that gives an AI coding agent real SEO data as callable tools. Instead of guessing search volumes, your agent calls the server and gets live keyword, SERP, backlink, audit and AI-visibility data back. You connect it once and every future session has it.
AI agents can write 2,000 words in thirty seconds.
They just do not know what to write about.
So they guess at keywords, invent search volumes, and cite competitors that never ranked.
omnifamous MCP fixes that.
Every SEO surface, as tools
The same engine that runs the omnifamous dashboard, exposed as tools your agent can call directly.
Keyword research
Real search volume, difficulty, CPC and intent from live data, not a number the model invented. Cluster by intent and rank by how winnable each one is.
SERP and rankings
Pull the actual top 100 for any keyword, in any country, on desktop or mobile. See who owns the page before your agent writes a word.
Competitors and content gaps
Which queries your rivals rank for and you do not. Domain intersections, ranked keyword tables, and the pages earning the traffic.
Backlinks
Referring domains, anchors, spam scores and broken links. Find the gap between your profile and the three sites beating you.
AI visibility
Run any prompt against ChatGPT, Gemini, Perplexity, Copilot, Grok or Google AI Overviews and get the answer, the citations and the query fan-out back.
Site audits
Technical crawl, Core Web Vitals, indexability, duplicate tags and the critical issues ranked by what actually moves rankings.
Article drafting
Brief, research and draft a full article against a target keyword, in your brand voice, with citations and internal links built in.
Calendar and publishing
Read and reshuffle your content calendar, then publish straight to Webflow, Ghost, WordPress and the rest of your stack.
AI visibility as a primitive, not a dashboard
Every other AI-visibility product hands your agent yesterday's stored numbers. These tools run the prompt live and hand back the answer, the citations and the query fan-out the engine actually issued.
answer, citations and the search fan-out it issued
the overview block for a live query
the full conversational search surface
answer and grounded sources
answer, sources and its search-model queries
answer and citations via Claude's own web search
answer and cited sources
answer and cited sources
ai_run_prompt(prompt: string, engine: string, country?: string)the full answer text, every citation with url and domain, the query fan-out the engine actually issued, the model that answered, and the real cost of the call
ai_run_prompt_multi(prompt: string, engines: string[], country?: string)the same payload per engine, side by side, so your agent can diff how ChatGPT and Gemini answer the same question
ai_check_brand(prompt: string, brand: string, competitors?: string[], engines?: string[])whether you were mentioned, where in the answer, in what sentiment, and who was named instead
ai_get_share_of_voice(workspaceId: string, window?: 7d|30d|90d)your share of answers against each tracked competitor, over time
ai_get_top_cited_domains(workspaceId: string, engine?: string, window?: string)the sources these engines actually pull from for your topics
ai_track_prompt(prompt: string, engines: string[], schedule?: daily|weekly)a tracked prompt that keeps scanning after the session ends
Ask an agent today how to track your brand in ChatGPT and it will tell you to build the probe loop yourself. These tools are that loop, already built, callable in one line.
Every tool, with its arguments
Free means it reads your own workspace. Metered means it buys live market data or creates something, and the call tells your agent the exact cost and your remaining balance.
Keywords
| Tool | Arguments | Returns | Cost |
|---|---|---|---|
seo_research_keywords | seed: string[], country?: string, language?: string, limit?: number | keyword, volume, difficulty, cpc, intent, trend[12] | metered |
seo_get_keyword_overview | keywords: string[], country?: string | volume, difficulty, cpc, competition, monthly_searches[] | metered |
seo_get_related_keywords | keyword: string, depth?: 1-4, limit?: number | related keywords with volume and difficulty | metered |
seo_list_keyword_pool | workspaceId: string, cursor?: string | your saved keywords, scores, assigned articles | free |
seo_check_keyword_coverage | keywords: string[] | which of these you already cover, and where | free |
SERP and competitors
| Tool | Arguments | Returns | Cost |
|---|---|---|---|
seo_get_serp | keyword: string, country?: string, device?: desktop|mobile, depth?: number | rank, url, title, description, SERP features, AI Overview presence | metered |
seo_find_serp_competitors | keywords: string[], country?: string | domains ranked by visibility across the keyword set | metered |
seo_get_domain_overview | domain: string, country?: string | organic traffic, keyword count, rank distribution | metered |
seo_get_content_gap | domain: string, competitors: string[] | keywords they rank for and you do not, scored | metered |
seo_list_competitors | workspaceId: string | your tracked competitors and their last snapshot | free |
Backlinks and audits
| Tool | Arguments | Returns | Cost |
|---|---|---|---|
seo_get_backlink_summary | domain: string | referring domains, backlinks, spam score, rank | metered |
seo_get_link_gap | domain: string, competitors: string[] | domains linking to them but not to you, by authority | metered |
seo_run_audit | url: string, depth?: number | job id; poll with seo_get_job | metered |
seo_get_audit_report | auditId: string | issues by severity, Core Web Vitals, indexability | free |
seo_get_search_console_performance | site: string, dateRange: string, dimensions?: string[] | clicks, impressions, CTR, position by query and page | free |
Content
| Tool | Arguments | Returns | Cost |
|---|---|---|---|
seo_draft_article | keyword: string, brief?: string, voice?: string | job id; the finished draft with citations and internal links | metered |
seo_list_content_calendar | workspaceId: string, from?: date, to?: date | scheduled slots, target keywords, status | free |
seo_get_article | articleId: string, format?: md|html|json | the article body, metadata and target keyword | free |
seo_publish_article | articleId: string, integrationId: string | the live URL on your CMS | free |
seo_get_brand_voice | workspaceId: string | the learned voice profile the drafter writes in | free |
Plus the visibility tools above, and the calendar, publishing, brand-voice and job-polling tools. 40+ in total at launch.
Ask anything. Get real data back.
Copy a prompt and run it in your agent once the server is connected.
One server, the whole stack
How it compares to the other SEO MCP servers
If you already pay for Ahrefs or Semrush, their server is the cheaper first move. Here is exactly where we are different.
| omnifamous | DataForSEO MCP | Ahrefs / Semrush MCP | Free GitHub servers | |
|---|---|---|---|---|
| Hosted, nothing to run locally | yes | self-host or npx | hosted | you run it |
| Bring your own provider account | not needed | DataForSEO account required | paid plan required | usually required |
| Raw SEO data (keywords, SERP, backlinks) | yes | yes, the deepest catalog | yes, their own index | varies, often one source |
| Run an ad-hoc prompt on an AI engine | yes | no | no | no |
| Citations returned with the answer | yes | no | no | no |
| Writes an article, not just reads data | yes | no | no | no |
| Publishes to your CMS | yes | no | no | no |
| Free tier without a card | 10,000 credits | no | no | yes, that is the point |
| Best if you want | an agent that does the work end to end | the widest raw data catalog | the index you already pay for | to read the source and self-host |
MCP server, skill, plugin or subagent?
These get used interchangeably and they are not the same thing. Only one of them can fetch a real search volume.
MCP server
A service your agent connects to. It brings live data and actions that the model cannot do on its own. Works in every MCP client, not just Claude. This page is about one of these.
Skill
A markdown file of instructions Claude loads when a task matches. It changes how the model behaves. It cannot fetch a search volume, because it is text, not a service.
Plugin
A bundle. One installable package that can contain skills, hooks, subagents and MCP servers together. A distribution format, not a capability.
Subagent
A second model instance with its own context and its own tool list. Useful for parallel work. It still needs an MCP server underneath to get real data.
Reading your own data is free
Credits only move when a tool creates something or buys live market data. Everything else is unmetered.
Your workspace
Articles, content calendar, keyword pool, competitors, visibility history, audit reports. Read and edit as much as you want, no credits spent.
New data and new content
Keyword research, SERP lookups, backlink profiles, audits, visibility scans and article drafts. You see the exact cost and your remaining balance on every call.
3,000,000 credits a month
On Pro, refreshed every billing period, plus 10,000 free credits the moment you sign up. Top up any time.
Questions, answered
Everything you need to know about how the engine does the work for you.
An SEO MCP server is a Model Context Protocol server that gives an AI coding agent real SEO data as callable tools. Instead of guessing search volumes, your agent calls the server and gets live keyword, SERP, backlink, audit and AI-visibility data back. You connect it once and every future session has it.
Connect the server to your agent once with a config block, then ask in plain language. The agent picks the tools: it might call a keyword tool, then a SERP tool, then a drafting tool, in one turn. You never call a tool by hand.
It depends on what you already pay for. DataForSEO's server has the widest raw data catalog. Ahrefs and Semrush make sense if you hold a seat there. The free GitHub servers are fine for a single lookup. We are the one that runs an ad-hoc prompt against an AI engine and returns the citations, and the one that writes and publishes rather than only reading.
Several, and they are on GitHub. They wrap one data source, you run them locally, and you bring your own provider key. Ours is hosted with 10,000 free credits and no card, so the difference is whether you want to operate a server or just use one.
It is a standard way to describe an API so a model can use it without custom glue. The tools declare their arguments and their return shape, the client discovers them at connect time, and the model decides which to call. The value is not the transport, it is that one config block works in every agent.
Yes, and that is the client we tune for. One claude mcp add command, or a .mcp.json you commit so your whole team gets the same tools. Cursor, Codex, VS Code, Windsurf, Zed, Cline, Gemini CLI, Hermes and n8n all get tested config too.
A skill is instructions, a plugin is a bundle, an MCP server is a service. Only the server can fetch a live search volume or run a prompt against ChatGPT, because only the server can make the call. Skills and plugins shape behaviour around it.
It can write, and it cannot measure. Ask it for a search volume and it will produce a confident, invented number. That is the exact gap an SEO MCP server closes: the model keeps the writing, the server supplies the facts.
ChatGPT, Google AI Overviews, Google AI Mode, Gemini, Perplexity, Claude, Copilot, Grok. You send a prompt, you get the answer, every citation with its URL, and the query fan-out the engine actually issued.
Reads of your own workspace are free and unmetered: your articles, your calendar, your keyword pool, your visibility history. Anything that creates content or buys live market data spends credits. Every account starts with 10,000 free credits.
Not for the connector flow, which signs you in through your browser. If you prefer a config file, create a secret key in Settings under MCP & API and paste it into the snippet.
That is what we are building toward. The server is designed to accept per-call payment so an agent can buy exactly what it needs, and to fall back to a free preview plus a checkout link when it cannot pay.
Still have questions? Talk to us
Your agent is guessing. It does not have to.
Give it live keyword, ranking and AI-visibility data from the same engine that runs omnifamous. Start with a free audit, no account and no card.
Run my free audit10,000 free credits when you create an account.Join the Omnifamous waitlist
Tell us about your site and we'll let you in when your spot opens up.