BLACK FRIDAY

50% off your first 2 months on every plan

Code RANKNOV26

Ends Monday 30 Nov, 23:59 CET

Claim the deal →

MCP server · Live since June 2026 · Updated November 2026

What can you do with theRankscale MCP server for AI Search?

Chief of Staff, RankscaleFacts checked against the product and the OpenAPI specification, November 2026

The Rankscale MCP server connects Claude, ChatGPT, Cursor and any other MCP client to your AI search visibility data, so you ask in plain language and get your dashboard's numbers with the AI answer behind them. As of November 2026, every plan includes it, from Essentials (€17 a month, billed yearly) up. It suits marketers, account managers and leadership.

What can the assistant reach? Everything in your dashboard, including each AI answer, and it can set up your tracking too:

  • 31 tools. 20 to read, 9 to edit your setup and 2 to run tracking. You choose which ones the assistant may use when you connect.
  • The full AI answer behind any number. The complete answer, every cited URL, every company it names and the AI Overview references, for any run.
  • 10 analytics views, not one score. Visibility, competitors, citations, AI source boxes, sentiment, shopping, query fan-out, ads, local and Maps results, and ChatGPT follow-up forms, each with filters for time frame, engine, topic, tags and query, plus a time series.
  • Changes your setup from a chat. Creates search terms, topics and dashboard views. New search terms start paused, and before anything spends credits the assistant shows the cost and waits for your confirm. It never deletes anything.
  • Works on 17+ AI engines, for every country and language you track, whatever plan you are on.

Need the data in your own reports, warehouse or automations? Use the Rankscale REST API: every raw AI answer as JSON, 11 reporting endpoints and bulk setup, from Pro+ (€229 a month).

Do you need the MCP server or the API?

Pick the MCP server when a person asks the questions, and the REST API when software does.

  • MCP server: you ask in Claude, ChatGPT, Cursor or OpenAI Codex, sign in with your Rankscale login and get the answer as a sentence, a list or a table. Included from Essentials (€17 a month, billed yearly).
  • REST API: your own code pulls JSON with an API key, for scheduled reports, a warehouse or alerts. From Pro+ (€229 a month).
  • Looker / Data Studio integrator: charts and scorecards in Looker Studio, no code and no chat. From Pro (€99 a month).

MCP server

Can AI assistants use Rankscale data?

Yes. The Rankscale MCP server is a remote MCP server that connects AI assistants such as Claude, ChatGPT, Cursor and OpenAI Codex to your Rankscale data, on every plan from Essentials (€17 a month, billed yearly). You sign in with your Rankscale account, so there is no API key to copy and nothing to install. The assistant reads the same data as your dashboard, and it can open the full AI answer and its sources behind any number. MCP (Model Context Protocol) is the open standard AI assistants use to connect to outside tools.

https://mcp.rankscale.ai/mcp

How do you connect Claude, ChatGPT or Cursor?

Add the server URL as a connector, sign in with Rankscale and choose what the assistant may do. Any client that supports remote MCP servers over Streamable HTTP with OAuth can connect.

Claude (web and desktop)

  1. Open Customize, then Connectors, then Add connector.
  2. Name it Rankscale and paste the server URL.
  3. Sign in to Rankscale, choose permissions, click Allow.
  4. Start a new chat. Claude uses the tools on its own.

ChatGPT

Needs a ChatGPT plan with developer mode.

  1. In Settings, turn on Developer mode.
  2. Add a new custom MCP plugin, paste the URL, set authentication to OAuth.
  3. Click Sign in with Rankscale and allow.
  4. In a chat, type @ or + to attach Rankscale.

Cursor, Codex and others

Add a remote server to your mcp.json. The first connection opens the same sign-in.

{ "mcpServers": {
  "rankscale": {
    "url": "https://mcp.rankscale.ai/mcp"
} } }

Step-by-step screenshots are in the help center article AI Assistants (MCP) and in the MCP docs. If your company uses an authenticator app, the sign-in asks for the second factor and then completes.

Can't find Rankscale in your assistant's connector list? You don't need to: add a custom connector and paste the server URL https://mcp.rankscale.ai/mcp. The steps above work in Claude, ChatGPT and Cursor.

What can you ask?

Ask about your own brand the way you would ask a colleague. The assistant picks the right tools, applies filters such as time frame, engine and topic, and answers with numbers that match your dashboard.

Questions about your visibility (read)

  • How is my brand trending across AI engines this month? Compare each engine with last month and name the engine that moved most.
  • Which competitors gained the most visibility in the last 30 days, on which engines and topics, and which of my search terms did they take?
  • Summarise sentiment for my brand over the last 30 days: what is positive, what is negative, and which sources are cited next to the negative mentions?
  • What are my top 10 citation sources right now, and which sites cite my competitors but never me?
  • List my search terms and flag the ones where I am not named in most runs, with the brand that is named instead.
  • Compare my shopping visibility this week versus last week, product by product, and tell me who wins the buybox now.
  • Show me the full answer ChatGPT gave for that search term in its latest run, which brands it named in which order, and which URLs it cited.
  • In which AI answers with a map of local places is our business missing, and which places are listed instead?

Changes to your setup (needs "Edit setup")

  • Create search terms for these five product pages on ChatGPT and Google AI Mode, tag them 'product-pages' and leave them paused.
  • Add a topic called 'Enterprise' and move every search term that mentions enterprise, SSO or security into it. List what you moved.
  • Rename the search term 'best crm' to 'best CRM for B2B companies' and add the tags 'decision' and 'crm'.
  • Save a dashboard view filtered to my top 3 competitors on Google AI Mode, and make it my default view.
  • Find the search terms where no brand was named in the last 30 days, list them, and pause them once I confirm.

Running tracking (needs "Run tracking", spends credits)

  • Run these five search terms now on ChatGPT so I can see today's answers. Tell me the credit cost first.
  • Start tracking these three search terms every week, and show me the credit cost per month before you switch them on.

Shows the credit cost first and waits for your confirm

What can the assistant change, and what does it cost?

Reading is always on and costs no credits; editing and running tracking are off until you switch them on. You choose when you connect, and every change asks you to confirm in the chat first. The assistant never deletes anything.

PermissionToolsWhat the assistant can doCredits
Read (always on)20Read your brands, search terms, topics, runs, saved dashboard views and every analytics summaryNone
Edit setup (off by default)9Create and edit search terms, topics and dashboard views. New search terms start paused, and creating them twice does not duplicate themNone
Run tracking (off by default)2Start recurring tracking on 1 to 50 search terms, or run 1 to 20 once nowYes. Cost shown first, waits for confirm
  • Only your own data. The tools return only what the signed-in user can already see in Rankscale, one question at a time. The assistant reads runs in pages, which suits analysis. To export every run into a warehouse or report, use the API's /executions endpoint.
  • Short-lived sign-in. OAuth 2.1 with PKCE. Access tokens last one hour and refresh on their own. Rankscale never issues long-lived keys for MCP.
  • Revoke at any time in Settings, Connections, MCP. It takes effect immediately. To add a permission later, disconnect and connect again.

What can the MCP server do?

The MCP server has 31 tools: 20 to read, 9 to edit setup and 2 to run tracking. Every analytics tool takes the same filters (time frame 24h, 7d, 30d, 3m or 1y; engine; topic; tags; query) and returns a summary plus an optional time series.

ToolWhat it returns or does
Read: setup and runs (no credits)
rankscale_brands_list, _getYour brands and their settings
rankscale_search_terms_list, _getYour search terms, with engine, schedule, tags and topic
rankscale_topics_listYour topics
rankscale_executions_listCompact summaries of runs, paginated
rankscale_executions_getOne run in full: the complete AI answer, cited URLs, companies mentioned and AI Overview references
rankscale_operation_getProgress of a run-now request
Read: analytics (no credits)
rankscale_dashboard_summaryVisibility Score, mentions, citations, sentiment, average position, Top 3
rankscale_competitors_summaryCompetitor leaderboard and each competitor's trend
rankscale_citations_summaryTop cited URLs and the sites that cite you most
rankscale_sources_summaryDomains shown in AI source boxes next to you
rankscale_sentiment_summarySentiment breakdown and the keywords driving it
rankscale_shopping_summaryProducts, merchants, buybox winners, price and rating signals
rankscale_fanout_summaryThe searches engines run before answering, and whether you are found
rankscale_ads_summaryAdvertisers and placements in AI answers
rankscale_local_results_summaryMap and place results: share, position, detection, top places, prompts where you are missing
rankscale_follow_up_forms_summaryChatGPT follow-up form rate, top questions and usual pre-selected answers
rankscale_custom_dashboard_views_list, _view_getYour saved dashboard views
Edit setup (no credits)
rankscale_search_terms_createCreates search terms, paused
rankscale_search_terms_updateChanges query, tags, schedule or engine
rankscale_search_terms_deactivatePauses 1 to 50 search terms after a preview
rankscale_topics_create, _assign_terms, _unassign_termsCreates topics and moves search terms between them
rankscale_custom_dashboard_view_create, _update, _set_defaultSaves and edits dashboard views and sets your default
Run tracking (spends credits)
rankscale_search_terms_activateStarts recurring tracking on 1 to 50 paused search terms
rankscale_search_terms_executeRuns 1 to 20 search terms once, now, without changing their schedule

What the MCP server does not do

  • It never deletes anything, and it cannot delete brands or your account.
  • It does not change brand or competitor settings. Do that in the dashboard or through the API.
  • The Rankscale MCP server reads the Visibility Score and each competitor's rank and detection rate, not share of voice, and says so instead of giving you one for the other. Share of voice exists only for ad placements.
  • One search term tracks one prompt on one engine. If you ask for several engines, it creates one search term per engine.
  • It cannot read Page Audit results or recommendations yet. Open those in the dashboard.

Which AI visibility platforms offer an MCP server, and what should you check?

Rankscale offers an MCP server on every plan, with 20 read tools and opt-in tools that set up and run tracking. Several AI visibility tools now offer an MCP server, and they differ on what the assistant may do, how you sign in, which engines it reads and which plan you need. Check these eight points for any tool, ours included.

What to checkWhy it mattersRankscale MCP server
Read only, or also set up tracking?Write tools let the assistant onboard prompts for you31 tools: 20 read, 9 edit setup, 2 run tracking
Can it open the AI answer behind a number?Without it you cannot check where a score comes fromYes: full answer, cited URLs, companies named
How do you sign in?An API key in a chat tool is a key that can leakYour Rankscale login, OAuth 2.1 with PKCE, no key
Who controls spending?An assistant that runs prompts spends creditsCost shown first, runs only after you confirm
Can it delete data?A wrong instruction should not wipe your setupNever deletes; new search terms start paused
Which engines does it read?Some tools lock engines to higher plansEvery engine you track, on every plan
Which plan do you need, monthly or yearly?MCP is sometimes enterprise-onlyEvery plan: from €17 a month billed yearly, or monthly from Pro (€99)
Which assistants connect?Your company may allow only one assistantClaude, ChatGPT, Cursor, OpenAI Codex, any remote MCP client with OAuth

What do teams do with the MCP server?

Teams run 9 jobs through the MCP server, from preparing a client call to finding why visibility dropped. We took them from sales and support conversations and from how our own team works in Claude. Copy a prompt, fill in the [brackets] and paste it into your assistant. Jobs that should run on a schedule, or feed a warehouse or a report, belong on the API: see the API use cases.

For agencies offering GEO and AI search services

Prepare a client call in two minutes

GEO and AI search agencies · MCP, read only

Before a client call, ask the assistant for the brand's trend, the competitors that moved and the sources the engines cite. You walk in with numbers that match the client's dashboard, and the AI answer behind any number you want to show.

  • Prepare me for a call with [client brand] in 30 minutes. From Rankscale, give me: the visibility score, mentions, citations and average position for the last 30 days against the 30 days before, per engine; the 3 competitors that gained most, and on which topics; the 5 domains cited most for the client's search terms, and whether the client's own site is among them; and one AI answer that shows the biggest win and one that shows the biggest loss, with the pages they cite. Finish with three talking points for the call.

Usesrankscale_dashboard_summaryrankscale_competitors_summaryrankscale_citations_summary

Check every client brand in one question

GEO and AI search agencies with many client brands · MCP, read only

On Monday morning, one question tells an account lead which client brands need attention this week. The assistant reads every brand you have access to and flags the ones under your threshold or falling, with the reason.

  • Go through every client brand in my Rankscale workspace and compare this week with last week. Which brands sit under a visibility score of 50, and which lost over 5 points? For each of those, tell me where it lost (engine and topic) and which search terms caused it. Biggest drop first, one line each.

Usesrankscale_brands_listrankscale_dashboard_summaryexample question from the OpenClaw guide

Set up a new client's tracking from a list

GEO and AI search agencies · MCP with "Edit setup"

Paste the client's prompts into the chat and let the assistant create the search terms and topics. New search terms start paused, so nothing runs and no credits are spent until you switch them on. One search term tracks one prompt on one engine, so asking for two engines creates two terms per prompt.

  • Here are 20 prompts for [client brand]: [paste the list]. Create one search term per prompt on ChatGPT and one on Google AI Mode, 40 in total, tagged 'onboarding-[month]'. Group them under three topics, Pricing, Comparison and How-to, by what each prompt asks. Leave every search term paused, then list what you created per topic so I can check it before anything runs.

Usesrankscale_search_terms_createrankscale_topics_createrankscale_topics_assign_terms

For in-house marketing and SEO teams

Answer leadership questions in the assistant they already use

In-house teams, CMOs, enterprise teams with an internal ChatGPT or Claude · MCP

The people who ask about AI visibility rarely log into a dashboard. With the MCP server connected to your company's Claude or ChatGPT, they ask in plain language and get the same numbers the dashboard shows, with the AI answer behind them.

  • Write a one-paragraph update for our CMO on how our brand did in AI answers this month: the visibility score and its change against last month per engine, the competitor that gained most and where, and the one AI answer that best shows where we lose. Use only numbers from Rankscale and name the search term behind that answer.
  • Which competitors gained the most visibility in the last 30 days, on which engines and topics, and which of our search terms did they take from us?
  • Show me the full answer ChatGPT gave for [search term] in its latest run, which brands it named in which order, and which URLs it cited.

Usesrankscale_dashboard_summaryrankscale_competitors_summaryrankscale_executions_get

Find out why visibility dropped

In-house SEO and content teams · MCP, read only

When a number falls, ask which engine and which prompts moved, then open the answers behind the drop. You see which brands and sources the engine named instead of you, in the same chat.

  • Our AI visibility dropped this week. Compare the last 7 days with the 7 days before, per engine and per topic, and tell me where the drop is. For the topic that fell most, list the search terms that lost us, and open the latest answer for each: which brands were named instead, and which pages were cited.
  • For those answers, what do the cited pages say that our site does not? Give me the 5 facts or questions we should add, and which of our pages each one belongs on.

Usesrankscale_dashboard_summaryrankscale_executions_listrankscale_executions_get

For content, SEO and PR teams

Build a content backlog from query fan-out

Content teams · MCP, read only

Before an AI engine answers, it runs its own searches. The fan-out view lists those searches and whether your site was found for each. The searches where you are never found are the questions your site does not answer yet: that is what to write or update next.

  • Using the Rankscale query fan-out data for the last 30 days, list the searches the AI engines ran before answering my tracked prompts, and whether my site was found for each. Keep the searches where my site was never found, on 2 or more engines. Group them by topic, count how many of my prompts triggered each search, and rank them by that count. For the top 10, tell me whether to extend an existing page (name it) or write a new one, and give me the question heading to add and the first sentence that answers it.

Usesrankscale_fanout_summary

Build a PR target list from citation gaps

PR and comms teams · MCP, read only

AI engines cite review sites, directories, media and forums. Ask which domains the engines cite for your prompts, for you and for your competitors. The sites that cite your competitors and not you are your outreach list, and sentiment shows which sources drive a negative description.

  • Using Rankscale citations and sources for the last 30 days, list the domains that AI engines cite for my competitors but never for my brand. For each domain, give the type of site (review platform, media, directory, forum or blog), how many answers cite it, on which engines, and which competitors it supports. Rank the top 15 as an outreach list, and mark any that are also cited next to negative mentions of my brand.
  • Summarise sentiment for my brand over the last 30 days: the share of positive, neutral and negative mentions per engine, the keywords behind the negative ones, and the sources cited next to them. Which source should we correct first, and what should we ask them to change?

Usesrankscale_citations_summaryrankscale_sources_summaryrankscale_sentiment_summary

For e-commerce, retail and multi-location brands

Ask where your products and stores show up

Online shops, retailers and brands with many stores · MCP, read only

Ask which products, merchants and prices the AI engines show and who wins the buybox, which advertisers appear in AI answers, and where an AI answer shows a map of local places without your business.

  • How did my products do in AI shopping answers this week compared with last week? Show which products appear, the shop and price shown for each, who wins the buybox, which rival products took our spots, and which advertisers show up on the same search terms.
  • When AI engines answer with a map of nearby places, where are we left out? Give me those search terms with the places that appear instead, their position and the engine, and tell me which of our stores and cities are missing most often.

Usesrankscale_shopping_summaryrankscale_ads_summaryrankscale_local_results_summary

For teams building their own AI agents

Give your team one AI visibility analyst in Claude

Teams that share a Claude workspace · MCP

An assistant that answers visibility questions for the whole team needs a clear brief: which data to use, what to quote and what it may change. Paste this brief as the instructions of a Claude Project with the Rankscale connector, so everyone gets the same analyst. Our webinar "From Manual Work to Automations" shows the MCP setup in Claude. For an agent that runs on a schedule with no one in the chat, use the API.

  • You are our AI visibility analyst. Use the Rankscale MCP tools to answer questions about our brand. For every number you report, open the AI answer behind it and quote the part that supports it. Never create, change or run search terms unless I ask in this chat.

Usesany MCP tool

Which plans include the MCP server?

Every Rankscale plan includes the MCP server, starting with Essentials at €17 a month, billed yearly ($17). Higher plans add credits, brand dashboards and data types for the assistant to read:

  • Essentials, €17 a month billed yearly: 120 credits a month and 2 brand dashboards.
  • Pro, €99 a month: 1,200 credits, 10 brand dashboards and unlimited seats, and it adds sentiment, shopping, local results, AI visits and ad analytics to your data.
  • Pro+, €229 a month: 3,000 credits and 25 brand dashboards, and adds the REST API.
  • Growth, €385 a month: 5,500 credits and 50 brand dashboards, with white-labeled share links.
  • Enterprise, €780 a month: 12,000 credits and 100 brand dashboards.

Dollar prices use the same figures ($99, $229, $385, $780), and yearly billing takes 15% off from Pro up. Prices as of November 2026; the free trial is 7 days on Pro. The pricing page has a credit calculator for your own prompts and engines.

Frequently asked questions

Which plans include the MCP server?

Do I need an API key for the MCP server?

No. You sign in with your Rankscale account (OAuth 2.1 with PKCE). Access tokens last one hour and refresh on their own, and you can disconnect at any time in Settings, Connections, MCP.

Should I use the MCP server or the API?

If a person asks the questions, use the MCP server. If code pulls the data on a schedule, use the API, from Pro+ (€229 a month).

Does asking the assistant cost credits?

No. Questions and reads are free. Credits go only to running search terms, and before the assistant runs one it shows you the cost and waits for your confirm.

Can an AI assistant change my setup or spend credits?

Only if you allow it. Editing and running tracking are off by default. New search terms start paused, and before anything spends credits the assistant shows the cost and waits for you to confirm. It never deletes anything.

Which AI assistants work with the MCP server?

Claude (web and desktop), ChatGPT with developer mode, Cursor, OpenAI Codex, and any client that supports remote MCP servers over Streamable HTTP with OAuth.

How can I see where a number in the chat comes from?

Ask the assistant to open the runs behind it. For any run it shows the full AI answer, every cited URL and every company named, so you see exactly where a number comes from. Ask how many runs sit behind a share as well: with only a few, one answer can move it a lot.

Does the assistant see the same numbers as my dashboard?

It does. The MCP server reads the same data as your dashboard and can open the full AI answer and its sources behind any number. It reads Visibility Score, not Share of Voice, and says so instead of swapping one for the other.

Can the MCP server export all my runs?

It reads runs page by page, which suits analysis in a chat. For a full export into a warehouse or report, use the API's executions endpoint, from Pro+ (€229 a month).

Can I use the MCP server on a monthly plan?

Yes, from Pro at €99 a month. Essentials (€17 a month) is billed yearly only. Every plan includes the MCP server.

Does it work with Microsoft Copilot?

No. Microsoft Copilot cannot connect to the Rankscale MCP server today. Use Claude, ChatGPT, Cursor or OpenAI Codex, or bring the data into Microsoft tools such as Power BI or Excel through the API or the CSV and Sheets exports (from Pro, €99 a month).


Changes: November 2026: rebuilt the MCP page. The REST API has its own page at rankscale.ai/api.

Sources (read 16 November 2026): Rankscale MCP docs, the help center article AI Assistants (MCP), the changelog (MCP server live since 11 June 2026), pricing and the webinar recording From Manual Work to Automations.

See also: Looker / Data Studio integrator · REST API · MCP docs · Pricing.