AI clients
CueScout in Claude Code
Claude Code already has your repo open, which is where the answers to most readiness findings live. Adding the CueScout server means the check and the fix happen in the same session: you ask whether the crawlers can reach the site, and the model that reads the answer is the one holding your robots.ts.
@cuescout/mcp is not on npm yet and the source is not public yet, so there is no way to run the server locally today. The config below is what you will paste when it lands. In the meantime the same five checks are available as public HTTP endpoints that need no account and no key — they are listed with each tool below, and an agent can call them directly.
Where the config lives: .mcp.json in the project root, or the `claude mcp add` command
Setting it up
1.Not installable yet
@cuescout/mcp is not on npm yet and the source is not public yet, so there is no way to run the server locally today. The config below is what you will paste when it lands. In the meantime the same five checks are available as public HTTP endpoints that need no account and no key — they are listed with each tool below, and an agent can call them directly.
What works today
curl -s https://cuescout.com/api/public/ai-readiness \ -H 'content-type: application/json' \ -d '{"url":"https://yoursite.com"}'2.Add the server
Run this from the project directory. Claude Code writes it to .mcp.json, so it is committed with the repo and every collaborator gets it.
In your project directory
claude mcp add cuescout -- node /absolute/path/to/cuescout/services/mcp-public/dist/index.js3.Check it registered
The server should appear with its five tools. If it does not, the usual cause is the path in the config — the client needs an absolute one.
In a Claude Code session
/mcp4.Ask it something worth asking
The most useful first question is not "is my site OK" but a specific buying question in your category, because the answer names whoever the engine recommends instead of you.
A prompt
Ask cuescout whether ChatGPT names us when someone asks "best tool for <your category>", then check whether our robots.txt lets its crawler in.
What this does not do
- The tools read public pages. They cannot see anything behind a login on the site you point them at.
- The visibility check queries a live engine, so the same question can come back differently an hour later. That is the engines being non-deterministic, not the tool being unreliable.
- Nothing is tracked between calls. Trend lines over time are the paid product, not the MCP server.
Frequently asked questions
Does the config get committed?
If you use `claude mcp add` from a project directory it writes .mcp.json in that project, which is meant to be committed. Use the user scope instead if you would rather keep it to your machine.
Does this need a CueScout account?
No. The five tools call public endpoints with no authentication — the same ones behind the free tool pages. There is no key to paste and nothing to sign up for.
What does it cost?
Nothing to you. Four of the five tools are HTTP fetches against the site you name. The visibility check puts a real question to a search-grounded engine, which costs us a model call, so it is rate-limited per IP.
Is any of this stored?
No. The public endpoints write nothing to a database. If you want the same checks tracked over time against a site you own, that is the paid product.
Related
Cursor
One block in ~/.cursor/mcp.json, and the agent editing your site can check what engines see of it.
AI clientsClaude Desktop
Paste one block into claude_desktop_config.json and ask about your site in ordinary chat.
AI clientsVS Code
A .vscode/mcp.json block — note the key is `servers`, not `mcpServers`.
Run the checks on your own site
The readiness check is free, needs no signup, and takes about half a minute. It works out what your site is built on and gives you the fixes written for it.