MCP server

The Rankry MCP server lets AI tools that speak the Model Context Protocol read your Rankry data. Ask Claude, ChatGPT, Cursor or another MCP client “How is my brand doing in AI search?” and it answers from your own reports, with the same numbers you see in the app.

What you see

Click Add integration in the sidebar and open MCP server. The page has two cards: API keys, where you create and revoke keys, and Connect a client, with the server URL and a setup snippet.

Create an API key

Click Create API key.
Under Key name, name it after the tool you’ll connect, for example “Claude Code”, and click Create key.
Copy the key straight away. Rankry shows it only once. Click Done when you’ve saved it.

Keys start with rnk_. The list shows each key’s name, a masked form with only its first and last characters, when it was created and when it was last used.

One key reads every project on your account. Make one key per tool, so you can revoke them separately.
You can have up to 20 active keys. Revoke one to make room for another.
To revoke a key, click its bin icon, then Confirm revoke. It stops working right away.

The MCP server counts as connected while you hold at least one active key. Your first key adds it to the sidebar; revoking the last one takes it out.

Connect a client

The server URL is:

https://rankry.ai/app/mcp

It uses Streamable HTTP and expects your key as a Bearer token in every request:

Authorization: Bearer rnk_xxxxxxxx

Your client needs to support remote (HTTP) MCP servers and let you set this header. The page has a snippet for two setups:

Config file
Claude Code

For Cursor, a project’s .mcp.json and other clients that read an MCP config file:

{
  "mcpServers": {
    "rankry": {
      "type": "http",
      "url": "https://rankry.ai/app/mcp",
      "headers": { "Authorization": "Bearer rnk_xxxxxxxx" }
    }
  }
}

Run it in a terminal:

claude mcp add --transport http rankry https://rankry.ai/app/mcp --header "Authorization: Bearer rnk_xxxxxxxx"

Right after you create a key, the snippet on the page already contains it, so you can copy it as is. Otherwise, replace the placeholder with your key.

Tools

Every tool takes an optional project. Without one, it uses your most recently updated project, so if you have several, ask your client to list your projects first.

Tool	What it returns
`list_projects`	Your projects, each with its current Rankry Score and last report date.
`list_reports`	A project’s reports, newest first, with score, status and dates.
`get_brand_summary`	The headline: Rankry Score and its change, visibility per engine, sentiment and top insights, with an optional trend.
`get_visibility`	Visibility per engine, with the categories where you’re strong, need work, weak or invisible.
`get_sentiment`	The positive, neutral and negative share of answers per engine.
`get_competitors`	A leaderboard of you and your tracked competitors: visibility, average position and sentiment.
`get_insights`	Insights from the latest report, each tagged with a severity and a status.
`get_history`	Rankry Score and per-engine visibility, sentiment, position and diversity over 7, 30, 90 or 365 days.
`get_top_prompts`	Your prompts ranked by how you perform: engines that named you, average rank, sentiment and competitors named. Latest report or a past one.
`list_sources`	The domains AI engines cite, ranked, with each domain’s type and Yours, Gap or Mixed status.
`get_ai_traffic`	Visits from AI answers: totals, visits per engine, landing pages, countries, devices and engagement.
`get_integrations`	The integrations on your plan and whether each is connected for a project.
`get_gsc_performance`	Search Console clicks, impressions, CTR and position by query, page, country, device, date or hour.
`get_gsc_index_status`	URL inspection for up to 25 URLs: indexed or not, coverage, robots.txt, last crawl and sitemaps.
`get_gsc_sitemaps`	Submitted sitemaps with status, errors, warnings and last download.
`get_pagespeed_reports`	Your saved PageSpeed audits: category scores and lab metrics per page and device.
`get_pagespeed_report`	One PageSpeed audit in detail: scores, metrics, field data and top failing audits.
`run_pagespeed_audit`	Runs a new PageSpeed audit with the project’s API key and saves the report.
`list_tasks`	Your tasks, filtered by status, type or title.
`get_task`	One task in full: summary, brief, steps, checklist and notes.
`get_guide`	How Rankry works, from its built-in guides.

Some tools need setup first: the Search Console tools need Search Console connected to the project, the PageSpeed tools need a PageSpeed key, and AI Traffic needs the tracking script and a plan that includes AI Traffic.

Access

The MCP server is on every plan, and keeps working in read-only mode. When your subscription has fully ended, requests are refused until you reactivate it.

Rankry Brain uses the same tools inside the app, limited to the project you’re in and without running audits.