Cituna MCP

Cituna MCP: the AI visibility MCP server

Cituna MCP is the official MCP server for [Cituna](https://cituna.com), the AI visibility tool: your AI agent reads how seven AI engines answer, mention and cite your brand every day, next to your Google Search Console data, and acts on it with 31 tools.

The seven engines are ChatGPT, Perplexity, Gemini, Claude, Grok, Google AI Overviews and Google AI Mode, and every tracked prompt is asked on all of them daily. Connect the server to Claude, ChatGPT, Cursor, VS Code or any other MCP client and you can ask which buyer questions you are losing, what each engine actually answered, which pages it read, and what to fix first, then have the agent fix it.

It is hosted, speaks Streamable HTTP, and signs you in with OAuth in your browser. One URL, nothing to install:

https://mcp.cituna.com/mcp

Running it yourself instead, for scripts, CI or a client without OAuth? The same server is on npm. It takes a personal API key and needs Node 20 or later:

npx -y cituna-mcp

Product page and full docs: [cituna.com/mcp](https://cituna.com/mcp).

Contents

• What your agent can do
• Try asking
• Quick start: claude.ai, Claude Desktop, Claude Code, ChatGPT, Cursor, VS Code, Windsurf, Codex CLI, Gemini CLI, any other client
• Tools
• Plans and access
• Built-in workflows
• What a session looks like
• Connecting Google Search Console
• Authentication and security
• Self-hosting the remote server
• Troubleshooting
• FAQ
• Links

What your agent can do

Most AI visibility tools hand you a dashboard and leave the thinking to you. Cituna puts the measurements where you already reason: ask your agent why an engine never mentions you, and it pulls the answer receipts, checks them against what Google already sends you, and queues the page that fixes it, without you opening a tab.

• Read the daily tracking grid. Which of your tracked prompts each engine cited you for today, at what position, and your current visibility score (get_visibility).
• Pull the receipts. What an engine actually answered for one prompt, the brands it named and the URLs it cited (get_engine_answers).
• Find where the answers come from. The third-party pages the engines cite for your prompts, how often, which rivals each one names, and whether you are on the page, read from the page itself rather than guessed (list_citation_sources).
• Spot new rivals. Brands the engines name in answers to your own buyer questions, and vendor names people search before reaching you, that you are not tracking yet (competitor_discovery).
• Join AI answers with Google. Search Console totals, any Search Analytics query, and six ready-made analyses: queries just off page one, pages seen and not clicked, demand by intent, pages never shown, your own pages competing, and searches worded like AI prompts (gsc_*).
• Work the fix queue. Every gap from your latest audit with its impact, effort and the concrete fix, and its status as you go (list_gaps, set_gap_status).
• Plan and ship content. Keywords with real demand and nothing written for them, the AutoSEO queue and every article it wrote, then queue, edit, refresh or skip from the chat (list_keywords, queue_article, update_article, enrich_article).
• Change what is measured. Add, disable or rewrite the buyer questions the seven engines are asked every day, with a before-and-after diff (set_prompts).
• Run a fresh scan when you have shipped a fix and want to look now (run_scan).
• Keep an outreach list. Mark a cited source as planned, pitched, placed or declined, so the list survives between sessions (set_outreach_status).

Try asking

Once connected, ask in plain English. The server tells your agent which tools and workflow answer which kind of question, so you do not have to name them.

Visibility and receipts

- "How is acme.com doing in AI search today? Which prompts and engines cite us?"

- "What did Perplexity answer for *best CRM for dentists*, and who did it cite instead of us?"

- "What changed in my AI visibility this week, and which engines dropped us?"

- "Why don't AI engines mention us? Use the not-cited diagnosis."

Sources and competitors

- "Which sites does each engine read for our prompts that don't mention us yet? Rank them as a pitch list."

- "Which of our own pages do the engines already cite?"

- "Who comes up instead of us, and which rivals aren't we tracking?"

- "Mark the G2 category page as pitched." *(Pro)*

Search Console, joined up

- "Which queries are just off page one, and which page ranks for each?"

- "Which pages get impressions and no clicks, and why?"

- "Where do Google impressions exist but AI never cites us? That is my AEO target list."

- "Do two of our pages compete for the same query?"

- "How many of our searches look like AI prompts?"

Fixes and content

- "List the open gaps for acme.com, ranked by the Search Console impressions each one touches."

- "Which keywords have demand and no page behind them? Queue the top three." *(Pro)*

- "Which AutoSEO articles are due a refresh? Refresh them." *(Pro)*

- "Add the three questions we lost this week to our tracked prompts, and disable the ones nobody searches." *(Pro)*

- "Run a fresh scan for acme.com and summarise the top 3 gaps." *(Pro, uses one scan)*

Quick start

You need a Cituna account on Starter or a higher plan, with at least one tracked brand. The 3-day free trial is app-only: MCP access starts when your plan does. A brand-new workspace has nothing to read yet, so add your brand in the app first (cituna.com).

There are two ways to connect, and both expose exactly the same tools:

claude.ai (web and mobile)

1. In Claude, open Customize → Connectors, click Add, then Add custom connector.
2. Name it Cituna and paste the URL:

   https://mcp.cituna.com/mcp

1. Claude sends you to Cituna to sign in, then shows a consent screen. Leave Also allow changes ticked to let Claude run scans and work your queue (Pro and Max), or untick it for a read-only connection.
2. In a chat, turn Cituna on from the + menu and ask your first question.

On Team and Enterprise plans an owner adds the connector once under Organization settings → Connectors, then each member connects it from Customize → Connectors and signs in with their own Cituna account. A connector added on claude.ai is also available in Claude Desktop and the mobile apps signed in to the same account.

Claude Desktop

The connector above is the quickest route: Customize → Connectors → Add → Add custom connector, paste https://mcp.cituna.com/mcp, and approve in the browser.

To run the npm package locally instead, generate an API key (below) and edit claude_desktop_config.json (Settings → Developer → Edit Config):

{

"mcpServers": { "cituna": { "command": "npx", "args": ["-y", "cituna-mcp"], "env": { "CITUNA_API_KEY": "cituna_sk_your_generated_key" } } } }

On Windows, if the server does not start, run npx through cmd: "command": "cmd" and "args": ["/c", "npx", "-y", "cituna-mcp"]. Restart Claude Desktop. The tools appear under the + button in the chat box.

Claude Code

Hosted, with browser sign-in:

claude mcp add --transport http cituna https://mcp.cituna.com/mcp

Then run /mcp inside Claude Code, pick cituna and choose Authenticate. To use an API key instead of OAuth:

claude mcp add --transport http cituna https://mcp.cituna.com/mcp \

--header "Authorization: Bearer cituna_sk_your_generated_key"

Or run the npm package locally, available in every project:

claude mcp add cituna --scope user \

--env CITUNA_API_KEY=cituna_sk_your_generated_key \ -- npx -y cituna-mcp

Run /mcp to confirm it is connected, then ask *"what changed in my AI visibility this week?"*

ChatGPT

ChatGPT connects to the same hosted URL. Yes, you can ask ChatGPT how ChatGPT cites you.

1. In ChatGPT on the web, open Settings → Security and login and turn on Developer mode. It comes with Plus, Pro, Business, Enterprise and Edu.
2. Open Plugins, click +, then Create MCP App.
3. Name it Cituna, paste https://mcp.cituna.com/mcp, choose OAuth, and create it.
4. Sign in to Cituna and approve with Also allow changes ticked if you want writes. Then add Cituna to a chat from the + menu.

ChatGPT asks before each change. Reading works on Starter; scans and edits need Pro or Max. Approved ChatGPT connections are listed under Integrations in the Cituna app, where you can disconnect them.

Cursor

Or add it to ~/.cursor/mcp.json (every project) or .cursor/mcp.json (one project):

{

"mcpServers": { "cituna": { "url": "https://mcp.cituna.com/mcp" } } }

Cursor lists the server under its MCP settings with a sign-in button; click it, sign in to Cituna and approve. To skip OAuth, send your API key as a header:

{

"mcpServers": { "cituna": { "url": "https://mcp.cituna.com/mcp", "headers": { "Authorization": "Bearer cituna_sk_your_generated_key" } } } }

Or run