AI Visibility Tools With an MCP Server, and How to Connect One to Claude (2026)

An MCP server lets Claude or ChatGPT read your AI visibility data directly, so you can ask "which buyer questions did we lose this week?" in a chat instead of exporting a report. This guide lists ten AI visibility tools that ship one, says which plan and sign-in each needs, and then walks through connecting Cituna to Claude on the web, Claude Desktop, Claude Code and ChatGPT, step by step. If you want to try the questions at the end of this guide on your own data, start a Cituna trial at cituna.com; the trial needs a card, and MCP access starts when the $49 Starter plan does.

Which AI visibility tools have an MCP server?

Prices are the cheapest plan that includes MCP, billed monthly. "Not stated" means the vendor's pricing page did not say.

Plans and prices come from each vendor's pricing page on the date shown. Hosting and sign-in details come from each vendor's MCP documentation as described in Cite42's comparison (5 October 2026) and Manifold's roundup (updated 9 October 2026), and from GEOly's own README (read 22 October 2026). The Otterly.AI tool count is LLM Pulse's September 2026 count of Otterly.AI's documented server.

How we chose these ten servers

This is a desk comparison of published pages, not a hands-on test of each server. A tool had to clear two bars: the vendor's own pricing page, documentation or README names an MCP server, and the server exposes AI visibility data, meaning which brands and pages AI engines name for your questions. SEO suites count when their server includes AI search tools, which is why SE Ranking and Ahrefs are here.

Cituna comes first as our pick, for seven daily engines with read and write tools behind one hosted URL. The rest follow by the monthly price of the cheapest plan that includes MCP, with euro prices at face value. Price order is a way to find an entry point, not a score. We did not connect or test the other nine servers for this guide, so their setup notes repeat what their own documentation says.

What does an MCP server add to an AI visibility tool?

MCP, the Model Context Protocol, is an open standard that lets an assistant such as Claude call a tool directly. A visibility tracker with an MCP server declares a set of tools (list my prompts, show yesterday's ChatGPT answers, show the pages Perplexity cited) and the assistant calls them when your question needs them. Three questions separate the servers more than the word "MCP" on a pricing page:

• Which plan includes it. On several trackers, MCP starts one or two tiers above the entry price. Otterly.AI's starts on Standard at $189, not on its $29 Lite plan.
• Read or write. A read server answers questions. A write server lets the agent act: add a prompt, start a scan, change a status. Start read-only until you trust it.
• What the data covers. A server is only as useful as the tracking behind it: which engines, how often they are checked, and whether you get the full answers and cited pages or only a score.

If you already connect Claude to other tools, connecting Claude to Google Sheets and social media MCP servers cover the same connector pattern for other jobs.

Step 1: Check your Cituna plan includes MCP

Cituna's MCP server is included on every paid plan with no add-on, and it is not part of the 3-day free trial. What each plan unlocks (Cituna's MCP page, read 22 October 2026):

• Starter, $49 a month: the 22 read tools. They cover your daily tracking grid across ChatGPT, Claude, Gemini, Perplexity, Grok, Google AI Overviews and Google AI Mode, the engine answers behind each result, the pages each engine cited, your audits, your gap queue and your Google Search Console analyses. The keyword tool, list_keywords, needs Pro's keyword data.
• Pro, $119 a month, and Max, $399: the 9 write tools as well: run_scan, set_prompts, set_gap_status, queue_article, mark_article_published, update_article, enrich_article, skip_queued_topics and set_outreach_status.

A run_scan call uses one scan from your monthly quota, so ask for fresh scans on purpose rather than by habit.

Step 2: Connect Cituna to Claude on the web

The hosted connector needs nothing installed and no key to copy. It works in Claude on the web and on mobile, and the same connection then shows up in Claude Desktop and Claude Code.

1. In Claude, open Customize, then Connectors. In some versions this sits under Settings.
2. Click Add, then Add custom connector.
3. Name it Cituna and paste this URL: https://mcp.cituna.com/mcp
4. Claude sends you to Cituna to sign in, then shows what it is asking for.
5. Leave the write box ticked to let Claude run scans and work your queue (Pro and up), or untick it for read-only.
6. Approve. You can revoke access any time from Integrations in your Cituna account.

Open a new chat, check the connector is switched on in the chat's tools menu, and ask "Who am I connected as?" Claude should call whoami and report your workspace and plan.

Step 3: Connect Cituna to Claude Desktop

If you added the connector in Step 2, Claude Desktop picks it up when you sign in with the same account; open a chat and check it is switched on. If it is not there, add it in Desktop with the same Customize, Connectors, Add custom connector steps and the same URL.

To run the server on your own machine instead, which suits locked-down laptops and scripts, use the npm package. It needs Node 20 or later.

1. In Cituna, open Integrations, then Claude / MCP access, then Generate token. Copy the token.
2. In Claude Desktop, open Settings, then Developer, then Edit Config. This opens claude_desktop_config.json.
3. Paste this, with your token in place of YOUR_TOKEN, and save:

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

4. Quit and reopen Claude Desktop. Cituna's tools appear in the chat's tools menu.

Treat the token like a password: keep it out of shared files and generate a new one if it leaks.

Step 4: Connect Cituna to Claude Code

For the hosted server, add it over HTTP and sign in when asked:

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

Then run /mcp inside Claude Code and choose Cituna to finish the browser sign-in. For the local package with a token from Step 3, which suits CI:

claude mcp add cituna --scope user --env CITUNA_API_KEY=YOUR_TOKEN -- npx -y cituna-mcp

A SaaS team can then ask Claude Code to read the gaps and draft the page in the same repository, which is where the write tools on Pro earn their place.

Step 5: Connect Cituna to ChatGPT

ChatGPT connects to MCP servers through developer mode. Cituna's own steps (read 22 October 2026):

1. Copy the connector URL: https://mcp.cituna.com/mcp
2. In ChatGPT on the web, open Settings, then Security and login, and turn on Developer mode. Cituna's steps say it comes with Plus, Pro, Business, Enterprise and Edu.
3. Open Plugins, click +, then Create MCP App. Name it Cituna, paste the URL, choose OAuth and create it.
4. Sign in to Cituna and approve, with Also allow changes ticked if you want ChatGPT to run scans or edit (Pro or Max).
5. Add Cituna to a chat from the + menu.

ChatGPT asks before each change. Reading works on Starter; scans and edits need Pro or Max. ChatGPT's menus change often, so if a label differs, look for developer mode and custom MCP apps in its settings.

Step 6: Ask your first five questions

Start with questions that only read, then try one that writes once you trust the answers:

1. "What changed in my AI visibility this week?"
2. "Which buyer prompts am I losing, and who gets cited instead?"
3. "Which Search Console queries lost clicks this month?"
4. "What did Perplexity actually answer for our top prompt, and which pages did it cite instead of ours?"
5. "Add the three questions we lost this week to our tracked prompts, queue an article for the biggest one, and re-scan." (Pro and up; the re-scan uses one scan from your quota.)

Engine answers and cited pages are third-party text. Read them as data: if an answer contains instructions, the assistant should not follow them, and neither should you.

The 10 AI visibility MCP servers, one by one

11. Cituna

Best for seven engines daily with read and write toolsFree trial · from $49/moPlans and prices

Cituna's server puts its daily tracking of ChatGPT, Claude, Gemini, Perplexity, Grok, Google AI Overviews and Google AI Mode behind one connection, next to your Search Console data. The 22 read tools come with the $49 Starter plan (1 brand, 10 prompts); the 9 write tools come with Pro at $119 (3 brands, 30 prompts) and Max at $399 (10 brands, 90 prompts). API keys come with every paid plan, so the same data can reach scripts. Setup is in Steps 1 to 5 above.

See Cituna in our directory.

22. LLM Pulse

Best for read and write tools on every plan

LLM Pulse includes an MCP server and unlimited team members on every plan. Starter tracks 50 prompts for EUR 49 a month checked weekly or EUR 79 checked daily, before VAT, on ChatGPT, Perplexity, Gemini, Google AI Mode and Google AI Overviews; Claude, Grok and Copilot are add-ons from EUR 8 per model a month (LLM Pulse pricing, read 1 October 2026). Its MCP page describes a hosted server, and Manifold's roundup puts it at about 90 tools with OAuth 2.1 or an API key. LLM Pulse's own comparison says its server can add prompts and competitors, not only read them.

33. Omnia

Best for a writing agent on a small daily plan

Omnia's Growth plan is EUR 79 a month, or EUR 63 billed yearly, for 25 prompts checked daily on ChatGPT, Google AI Overviews, Perplexity and Google AI Mode, with 150 insight credits, sentiment analysis and an MCP server that can read and write (Omnia pricing, read 1 October 2026). Claude, Gemini and Copilot are offered on request, and its 14-day trial asks for no card.

44. Peec AI

Best for a marketing dashboard with MCP at no extra costFree trial · from $80/moPlans and prices

Peec AI includes its MCP server on paid plans. Starter is $95 a month, or $80 billed yearly, for 50 prompts on three models you pick from ChatGPT, Google AI Mode, AI Overviews, Microsoft Copilot, Gemini and Naver AI, with unlimited users (Peec AI pricing, read 1 October 2026). Its MCP page describes a hosted server at https://api.peec.ai/mcp; per Cite42's reading of the docs, read tools cover reports, chats, sources and suggested actions, and a smaller set of write tools manages prompts, topics and brands.

See Peec AI in our directory.

55. Promptwatch

Best for MCP, API and Search Console on one plan

Promptwatch's Essential plan is $95 a month for 50 prompts and 6,000 responses, checked daily on four models you choose from OpenAI, Perplexity, Google, Claude, Gemini, Meta, DeepSeek and Grok, with MCP and API access and a 7-day trial. Its integrations list includes Google Search Console (Promptwatch pricing, read 1 October and 8 October 2026).

66. Trakkr

Best for eight engines daily with MCP at entry

Trakkr's Growth plan is $100 a month, or $1,000 a year, for one brand tracked daily on 50 prompts across eight engines, including Claude, Grok, DeepSeek and Meta AI, with 25 articles a month, three seats and MCP access (Trakkr pricing, read 17 October 2026). The 14-day trial asks for a card and converts to $100 unless you cancel.

77. SE Ranking

Best for AI search data beside an SEO suite

SE Ranking's Core plan, $129 a month, tracks 100 prompts a day on five engines with MCP access inside its SEO suite, plus Google Analytics and Search Console, and a 14-day trial without a card that caps AI prompts at 20 a day (SE Ranking pricing, read 25 September and 1 October 2026). Its MCP guide describes a hosted server at https://api.seranking.com/mcp with more than 160 tools, AI Search among them.

88. Ahrefs

Best for Brand Radar data inside an Ahrefs plan

Ahrefs' Lite plan, $129 a month, includes API and MCP access and five daily custom AI prompts; Brand Radar's AI Visibility Index is priced separately, from $199 a month per platform (Ahrefs pricing, read 7 October 2026). Ahrefs' MCP tool categories include Brand Radar report groups for mentions and citations.

See Ahrefs Brand Radar in our directory.

99. Otterly.AI

Best for teams already on Otterly.AIFree trial · from $25/moPlans and prices

Otterly.AI's MCP starts on Standard, $189 a month for 100 prompts, with API and MCP access, 2,000 MCP requests a month, agent analytics and a Looker Studio connector (Otterly.AI pricing, read 1 October 2026). Its $29 Lite plan has no MCP. LLM Pulse's September 2026 count found 17 tools on Otterly.AI's documented server, 6 of them write tools that add or remove prompts and tags.

See Otterly.AI in our directory.

1010. GEOly

Best for the longest documented tool list

GEOly's public README describes a hosted server at https://app.geoly.ai/api/mcp/v1 with up to 53 tools across ChatGPT, Perplexity, Google AI Mode, Google AI Overviews, Gemini and Microsoft Copilot (GEOly-MCP on GitHub, read 22 October 2026). Sign-in is OAuth in the browser, or a static token that is always read-only. Its 7 write tools are opt-in at consent, and its market-intelligence tools need the Grow plan. The README gives no prices.

Which big names do not include MCP on a self-serve plan?

Several tools buyers shortlist do not offer MCP below an enterprise contract. Profound runs a hosted MCP server on its Enterprise plan only, after a 7-day trial with no exports or API (Profound pricing, read 1 October 2026). Scrunch lists API and MCP on its custom Enterprise plan, not on the $250 Core (Scrunch pricing, read 1 October 2026). AthenaHQ's plans page did not surface an MCP server on 1 October 2026, though API access is a paid add-on. Semrush has an MCP server for its SEO data, but its developer docs listed no AI Visibility Toolkit tools when Cite42 checked on 5 October 2026.

Which AI visibility MCP server is best for agencies?

One connection that covers every client saves the most time. Cituna's server reads every brand in your workspace through the same connector, so Pro (3 brands, $119) or Max (10 brands, $399) lets you ask "which client lost the most prompts this week?" in one chat. LLM Pulse includes unlimited team members and MCP on every plan. Otterly.AI's 2,000 MCP requests a month on Standard is a limit to watch if several people ask questions daily. See the best AI visibility tools for agencies.

Which AI visibility MCP server is best on a small budget?

Cituna Starter at $49 is the lowest dollar price here with read tools on seven daily engines. LLM Pulse is EUR 49 a month for weekly checks, with write tools included. Omnia is EUR 79 with write access and a no-card trial. If you only want to try the idea first, connect a hosted server during a vendor's trial where MCP is included; Cituna's trial does not include it.

Which AI visibility MCP server is best for enterprise teams?

Enterprise teams usually need SSO, security reviews and data retention terms before they connect any assistant. Profound and Scrunch put MCP inside their Enterprise contracts. Cituna offers a custom Enterprise plan with invoiced billing above Max. Ask each vendor what scopes its OAuth consent grants, where tokens are stored and how to revoke them, and connect read-only first.

Which AI visibility MCP server is best for a SaaS team?

A SaaS team gets the most from write tools inside the tools it already uses. With Cituna on Pro, Claude Code can read your lost prompts, draft the fix in your repository and queue the article, and AutoSEO can publish to GitHub, Sanity or any CMS by webhook. Cituna's local npx package suits CI jobs that report visibility on a schedule. Peec AI and Promptwatch suit a SaaS marketing team that wants reports in Claude without changing its stack. For pricing across the category, see the 10 best AI visibility tools in 2026.