quattr-ai-search · private beta
The AI visibility plugin for AI SEO inside Claude and ChatGPT
39 saved expert workflows for AI search, answer engine optimization, generative engine optimization and agentic search optimization (ASO), that your assistant follows step by step. The AI visibility plugin is an SEO plugin for Claude and ChatGPT: it carries the method, and the bundled MCP carries the 58 governed analyses it runs against your own data.
Scope resolved before filtering. Significance before any causal wording. A stated reason when it declines. Every skill, every time, that is what a saved expert workflow means here, as opposed to a prompt someone pasted into a doc.
Private beta The plugin is rolling out org by org. Request access ↗ , the steps below work once your org is in.
Install it
Claude & Claude Code
The plugin installs, and the MCP comes with it.
- Private beta: your org gets the plugin through workspace plugins, we set it up with you when you join.
- Sign in with your Quattr account when prompted, OAuth 2.1, read-only scope.
- Ask your first Quattr question. The bundled MCP routes it, no tool names, no setup beyond sign-in.
Works in Claude (web and desktop), Claude Code and Cowork. Re-authenticate any time with /mcp.
ChatGPT & Codex
The same skills, via the Agent Plugin standard, connector included.
- Private beta: your org gets the plugin through workspace plugins, we set it up with you when you join.
- Sign in with your Quattr account when prompted, OAuth 2.1, read-only scope.
- Ask your first Quattr question. The same skills route it, no tool names, no setup beyond sign-in.
Endpoint is the same on every client: https://mcp.quattr.com/mcp
What differs by client
The plugin follows the Agent Plugin standard, so the skill files install in Claude runtimes and in ChatGPT and Codex alike, same skills, same 58 analyses, same scope, freshness and refusal behavior. In MCP-only clients (Cursor, VS Code, Goose) you connect the MCP directly and the prompt gallery stands in for the skill layer, same analyses and guardrails; what you do not get there is the assistant following a saved workflow without being asked.
The 39 skills
Grouped by area. Each one names what fires it, what it runs, and, the part worth reading, what it refuses to do.
Organic 8
- The 30-second verdict: up, down, does it matter. Say: "how's search doing" 3 analyses
- The Monday report, written for you. Say: "weekly search report" 6 analyses
- The fastest ranking wins available now. Say: "quick wins this week" 2 analyses
- Positions that underearn their clicks. Say: "CTR opportunities" 2 analyses
- Demand your competitors own, you don't. Say: "keyword gaps vs rivals" 3 analyses
- Pages competing with themselves. Say: "are pages cannibalizing" 2 analyses
- AI Overviews & features on your terms. Say: "SERP feature changes" 2 analyses
- What moved, gained, and slipped. Say: "rank movement this month" 3 analyses
Market share 2
AI visibility 8
- The full answer-engine audit. Say: "run an AEO audit" 8 analyses
- Citation rate, trend, and gaps. Say: "how's our AI visibility" 2 analyses
- Watch citations move engine by engine. Say: "monitor our citations" 3 analyses
- Are AI bots fetching what you want cited? Say: "are AI crawlers reaching us" 4 analyses
- Quick per-bot fetch check. Say: "is GPTBot crawling us" reads rather than queries
- Prompts where you never appear. Say: "where are we invisible" 3 analyses
- Organic strength vs AI-answer presence. Say: "AI vs traditional gap" 4 analyses
- Why a page does or doesn't get cited. Say: "is this page citable" reads rather than queries
Content 2
Technical 4
- Did the algorithm update touch you? Say: "did the update hit us" reads rather than queries
- Crawl, index, and CWV in one sweep. Say: "technical health check" 7 analyses
- Core Web Vitals regressions, watched. Say: "any CWV regressions" 2 analyses
- Crawl budget going where it matters? Say: "crawl efficiency" 3 analyses
Paid 2
Revenue 1
Stats 3
Reporting 3
Utilities 6
- Ask Quattr anything about your own setup. Say: "what can Quattr do here" 15 analyses
- Scaffold a new skill that passes the gates. Say: "make me a custom skill" 2 analyses
- Pin segment, dates, and filters once. Say: "set my scope" 3 analyses
- Teach it your business context. Say: "build my context doc" 4 analyses
- First-session setup, guided. Say: "get me set up" 6 analyses
- When something looks wrong, diagnose. Say: "something looks off" 6 analyses
Questions people ask first
Is the MCP included with the plugin?
Yes. Installing the plugin in a Claude runtime brings the MCP with it, the 39 skills are the method, the 58 analyses are what they run. ChatGPT and Codex install the same skills through the Agent Plugin standard, connector included; MCP-only clients connect the MCP directly and use the prompt gallery in place of the skill layer.
Do I need to learn tool names?
No, and the design assumes you will not. You ask in your own words; the assistant reads the question and picks the analysis. When a question is too broad to pick from, it says which analyses it is choosing between and asks rather than guessing.
Which clients are supported?
Claude on web and desktop, Claude Code, Cowork, ChatGPT, Codex, Cursor, VS Code and Goose. The skill files install in the Claude runtimes and in ChatGPT/Codex via the Agent Plugin standard; MCP-only clients get the analyses plus the prompt packs, see "What differs by client" above.
Is my data safe?
The connection is OAuth 2.1 with a read-only scope, and your Quattr account decides what is visible, the assistant cannot see more than you can. There are no API keys to paste and nothing to store locally. Nothing is written back to your analytics or ads accounts.
What does it cost?
The plugin is in private beta, request access and we set your org up. The MCP is included with a Quattr subscription, with fair-use limits and no per-call metering, so exploring a question costs nothing extra, which is the point of asking follow-ups.
What happens if a source is not connected?
It says so, names the source, and stops. It does not approximate the answer from a source that cannot support it, a paid-and-organic question without Google Ads connected gets told that, not a half-answer built from organic alone.
Two minutes to connect
OAuth 2.1 sign-in, read-only scope, your data stays yours.
Request a demo Take a test drive → See it answering something first →