An AI visibility MCP is a server that lets your AI agent check whether ChatGPT, Gemini, Claude and Perplexity name your brand, and who they name instead. ContextBolt SEO is one: connect it to Claude, Cursor or Codex, ask “does ChatGPT recommend us for project management software?”, and the agent asks ChatGPT and reads you the answer with its sources.
MCP, the Model Context Protocol, is the open standard that lets an assistant use another product’s tools. If the term is new, our plain explainer on MCP covers it.
ContextBolt SEO is ours, so weigh the bias. On November 14, 2026 we used it to put this page’s own question to all four engines: what is the best AI visibility tool with an MCP server I can use in Claude? In the answers we saved, they named 16 different tools and agreed on one. None of the four named us.
- It answers one question from inside your agent: does ChatGPT, Gemini, Claude or Perplexity name you for what your buyers ask, and who does it name instead.
- Two kinds, priced two ways: trackers run a fixed prompt list on a schedule, such as 50 prompts for €49 a month at LLM Pulse or $95 at Peec AI. Live checkers charge per answer: ContextBolt SEO uses 2 credits an engine, after a 7-day free trial.
- The engines disagree: asked the same question on one day, the four named 16 different tools, and LLM Pulse was the one name two of them shared.
- Counted and live answers disagree too: LLM Pulse held 3% of counted mentions among five AI visibility brands, yet ChatGPT and Perplexity both put it first live that day.
- Setup is a URL and a sign-in on hosted servers like ours, Peec AI’s and Otterly’s.
What an AI visibility MCP does
An MCP server publishes a short list of tools, and your agent decides when to call them. AI visibility servers cover four jobs between them:
- Ask an engine live: Put one question to ChatGPT, Gemini, Claude or Perplexity and get back the words it wrote, the sources it cited and whether your brand is in there. In ContextBolt SEO that is
ai_answer, 2 credits an engine. - Count mentions: How often a brand turns up across a large set of past answers, by platform. Ours is
ai_visibility, 8 credits, counted across ChatGPT and Google’s AI answers. - Compare against rivals: Your share of those mentions next to up to four competitors. That is
ai_share_of_voice, 8 credits. - Read tracked prompts: A tracker runs your saved prompt list every day or week, and its MCP lets the agent read the results and trends.
A live answer has three outcomes. You are named, which is the win. You are not named but your page is cited as a source, which means the engine trusts your content and recommends someone else; cited but not recommended covers what to do about it. Or you are absent from both.
Every check in ContextBolt SEO also saves itself to your private SEO Dashboard, so next month’s run can lead with what moved.
We asked four engines which AI visibility MCP to use
From Claude Code on November 14, 2026, we sent ContextBolt SEO’s ai_answer the question “What is the best AI visibility tool with an MCP server I can use in Claude?”, United States, once per engine, watching for contextbolt.com. Four engines at 2 credits each came to 8 credits. Claude and Perplexity timed out at first, which cost nothing, and answered on a retry. ChatGPT’s and Perplexity’s answers ran longer than the copy our tool keeps, so the table lists the tools named before the cut.
| Engine | Tools it named | Sources cited | ContextBolt |
|---|---|---|---|
| ChatGPT | LLM Pulse, Peec AI, Otterly.AI, Rankscale (the first four of its top five) | 7 | Not named, not cited |
| Gemini | GoVISIBLE, SEOmatic, ZeroRank AI, Quattr | 7 | Not named, not cited |
| Claude | Firemark, Finseo, Cituna, Mentionable, GEO Optimizer | 5 | Not named, not cited |
| Perplexity | LLM Pulse, Profound, Visiblie, Aiso | 14 | Not named, not cited |
Asking one engine would have hidden most of this.
- Sixteen tools, one shared: ChatGPT and Perplexity both named LLM Pulse, and no other tool appeared twice. A buyer who asks Gemini gets a different shortlist from one who asks ChatGPT, so check every engine your buyers use.
- We were missing before the answer was written. ChatGPT searched 19 sites, Claude 10 and Perplexity 14 for this question, and contextbolt.com was in none of those searches. An engine cannot cite a page its search never found, so the fix starts with the pages those searches return.
- Each answer is one run. The same question can come back different an hour later, so a single miss is a reading for that run. Read the pattern over several checks.
The run is also why this page exists. None of the four found a page of ours to cite for the question.
Counted mentions and live answers disagree
A live answer reads today’s web. A mention count looks back over a large set of answers already collected. They measure different things, and the gap between them is useful.
The same day, we ran ai_share_of_voice on the topic “ai visibility” for five brands that sell AI visibility tracking, by name. It counts mentions across ChatGPT and Google’s AI answers.
| Brand | Mentions | Share |
|---|---|---|
| Profound | 316 | 54% |
| Otterly | 110 | 19% |
| Peec AI | 104 | 18% |
| Scrunch | 36 | 6% |
| LLM Pulse | 16 | 3% |
LLM Pulse holds 3% of the counted mentions, yet ChatGPT and Perplexity both put it first when asked live. A newer brand can win today’s answers before the counted history catches up, and an established one can lead the history while live answers move on. You want both readings before you decide where to spend effort.
One lesson from our own runs: pass names, not web addresses. Our first try used addresses, and the check counts an address by its first word, so tryprofound.com came back as “tryprofound” with 9 mentions and llmpulse.ai as “llmpulse” with 2. With the names people use, Profound counted 316 and LLM Pulse 16. Profound is also an everyday word, so part of its count may be the word rather than the company. Each run took 8 credits. What AI share of voice measures explains how to read the percentages.
AI visibility MCP servers compared
We read each vendor’s own MCP and pricing pages on November 14, 2026. Prices are as each vendor shows them; Otterly and LLM Pulse quote in euros.
| Server | How it gets answers | Engines | Cheapest way in with MCP |
|---|---|---|---|
| ContextBolt SEO | Asks live when you ask, plus counted mentions | Live: ChatGPT, Gemini, Claude, Perplexity. Counted: ChatGPT, Google AI | 7-day free trial, then $35 a month for 1,000 credits; 2 credits an engine per answer |
| LLM Pulse | Tracked prompts, weekly on the €49 plan | ChatGPT, Perplexity, Gemini, AI Overviews, AI Mode; Claude as an add-on | €49 a month for 50 prompts; 14-day trial |
| Peec AI | Tracked prompts, daily | Three on Starter, picked from ChatGPT, Perplexity, Gemini, AI Overviews, AI Mode and others | $95 a month for 50 prompts; MCP on every plan |
| Otterly.AI | Tracked prompts | Varies by plan | Standard, €189 a month for 120 prompts and 2,000 MCP requests |
| Ahrefs Brand Radar | Collected answers, plus custom prompts | AI Overviews, AI Mode, ChatGPT, Perplexity, Gemini, Copilot | Lite at $129 a month, plus Brand Radar AI from $199 a month |
| Profound | Tracked prompts | Up to 10 on Enterprise | Enterprise plan, price on request |
| Limelit | Credits per answer; runs on demand or on a schedule | ChatGPT, Perplexity, Claude, Gemini, AI Overviews, AI Mode | 1,000 free credits, then packs from $19.99 |
| Apify AI Visibility Tracker | Asks live per run, can be scheduled | ChatGPT, Gemini, Perplexity, Claude, AI Overviews, AI Mode | $0.04 to $0.06 a standard answer, $0.10 to $0.20 for Claude |
ContextBolt SEO is first because a live answer, a mention count and the SEO work that follows all sit on one $35 plan. When an engine skips you, the same agent pulls the keywords, the Google top 10 and your Search Console to fix the page it should have cited. It does not run prompts on a schedule: you ask when you need an answer, and a weekly check is a prompt you run. The trial runs 7 days with 100 credits, card required, and ContextBolt SEO vs Peec AI sets it against a daily tracker.
LLM Pulse lists its MCP server under the Starter plan on its pricing page: €49 a month for 50 prompts tracked weekly, or €79 for daily. The base engines are ChatGPT, Perplexity, Google AI Overviews, AI Mode and Gemini; Claude, Copilot and Grok are add-ons. The 14-day trial needs a payment method.
Peec AI shows MCP as included on every plan on its pricing page, from $95 a month for 50 prompts on three engines. The server is https://api.peec.ai/mcp with OAuth.
Otterly.AI leaves MCP off its €29 Lite plan. Standard at €189 a month adds it, with 2,000 MCP requests a month, per its pricing page.
Ahrefs puts MCP on every plan from Lite at $129 a month, and its MCP reaches Brand Radar, the AI visibility product. Brand Radar’s AI Visibility Index starts at $199 a month and prompt tracking at $50, per the Brand Radar page, so the AI data costs more than the plan that carries the server.
Profound’s MCP docs say the server is “Available on: the Enterprise plan”, which is priced on request.
Limelit starts new accounts with 1,000 credits and no card, and says MCP is free with any account on limelit.co. Packs run from $19.99, which it puts at about 1,100 answers a month.
The Apify actor is a third-party script hosted on Apify, which you reach through Apify’s own MCP server. Its actor page charges per answer and says failed checks are free.
Semrush sells an AI Visibility Toolkit and runs an MCP server, but we could not confirm from its MCP docs that the server returns the toolkit’s data, so it is not in the table. For trackers judged as dashboards, with or without MCP, see the best AI visibility tools.
What a check costs
Trackers price by prompts you keep. Live checkers price by answers you ask for. Which is cheaper depends on how you work.
- ContextBolt SEO: 1,000 credits for $35 is about 3.5 cents a credit. One engine’s live answer is 2 credits, about 7 cents. Asking all four engines one question is 8 credits, about 28 cents. A mention count or a share of voice check is 8 credits each. The trial’s 100 credits cover 50 live answers.
- Peec AI: 50 prompts on three engines, run daily, is about 4,500 answers a month for $95, about 2 cents an answer, as long as those 50 are the questions you want.
- Apify actor: $0.04 a standard answer on Apify’s $19 Starter plan, $0.06 on its free plan, and $0.10 a Claude answer on Starter.
- Limelit: one credit is $0.01 of usage, and its $19.99 pack is listed at about 1,100 answers.
A tracker wins when you want the same 50 questions asked every day and a history for each. Per-answer pricing wins when the questions change: before you write a page, after a launch, when a client asks why a rival keeps showing up. You pay for that one question, on the engines that matter, and skip the 49 you did not need that week.
How to connect an AI visibility MCP
Hosted servers connect with a URL and then a sign-in. These are the lines for ContextBolt SEO, from our setup page.
- Start the trial. Open ContextBolt SEO and start the 7-day free trial. The server URL is
https://seo.contextbolt.app/mcpfor every account. - Add it to your agent. Pick your client below.
- Sign in. Approve the connection in the browser window that opens.
- Check it worked. Ask “how many SEO credits do I have left?” The credit check is free and proves the connection.
Claude Code:
claude mcp add --transport http contextbolt-seo https://seo.contextbolt.app/mcp
Then run /mcp and choose Authenticate.
Claude Desktop: open Customize, then Connectors, then +, then Add custom connector. Paste the URL, leave the options Claude fills in as they are, and sign in.
Cursor: add the server to mcp.json, then click Authenticate.
{
"mcpServers": {
"contextbolt-seo": {
"url": "https://seo.contextbolt.app/mcp"
}
}
}
Codex:
codex mcp add contextbolt-seo --url https://seo.contextbolt.app/mcp
codex mcp login contextbolt-seo
The trackers above swap in their own URL, such as https://api.peec.ai/mcp or https://data.otterly.ai/mcp, and the same sign-in step.
Prompts that get usable answers
The server supplies the tools. The prompt decides what you learn and what it costs.
- Ask as a buyer would. “Ask ChatGPT, Gemini, Claude and Perplexity: what is the best CRM for a two-person agency? Tell me who each one names and whether we appear.” A bare brand name tells you little; a buyer’s question shows who wins the sale.
- Follow the sources. “When it skipped us, which pages did it cite?” Those pages are your list of places to get mentioned or to beat with a better page.
- Compare once a month. “Compare our share of AI mentions for small business accounting with Xero and FreshBooks.” One 8-credit check gives you the leaderboard.
- Ask again later. “Run the same question on ChatGPT again.” A repeat check costs its credits and opens with what changed since the last one.
If you would rather try before connecting anything, our free AI visibility checker runs two checks on ChatGPT and Google AI with no sign-up. For a full routine from prompt list to fixes, AI search visibility with Claude walks through it, and keyword research MCP servers covers the keyword side of the same agent.
Which AI visibility MCP to pick
Find your situation, then read across.
| Your situation | Pick |
|---|---|
| You do your own SEO in Claude, Cursor or Codex and want to ask the engines when a question comes up | ContextBolt SEO |
| You want the fix in the same chat as the finding: keywords, the top 10, your Search Console | ContextBolt SEO |
| You want the same prompts tracked every day, in a dashboard | Peec AI, or LLM Pulse for less with weekly tracking |
| You already pay for Ahrefs and want AI data next to it | Ahrefs with Brand Radar |
| You want to pay per answer with no subscription | Limelit or the Apify actor |
| You report on many brands for an enterprise team | Profound |
Skip ContextBolt SEO if your job is a fixed prompt set tracked every day across engines like Copilot or Grok: it asks four engines live and does not run on a schedule. If your job is finding out why AI names a rival and then fixing the page, it is the tool we would pick, and we build it.
Start the 7-day free trial of ContextBolt SEO, connect it with the line for your client above, and ask the four engines the question your best customer typed before they found you.