Connect verified.fyi to Claude, ChatGPT, Cursor or any app that supports MCP. Before your assistant buys from, books with or signs up to a site, it can read the same trust report you'd see here.
https://verified.fyi/mcp
https://verified.fyi/mcp as the URL. Leave the OAuth fields empty.Run this in your terminal. Add --scope user after add to use it in every project.
claude mcp add --transport http verified-fyi https://verified.fyi/mcp
https://verified.fyi/mcp as the server URL and set authentication to No authentication.Or add this to ~/.cursor/mcp.json, or .cursor/mcp.json in a project.
{
"mcpServers": {
"verified-fyi": {
"url": "https://verified.fyi/mcp"
}
}
}
Or add this to .vscode/mcp.json, then use it from Copilot Chat in agent mode.
{
"servers": {
"verified-fyi": {
"type": "http",
"url": "https://verified.fyi/mcp"
}
}
}
Run this in your terminal.
gemini mcp add --transport http verified-fyi https://verified.fyi/mcp
Any app that connects to remote MCP servers over Streamable HTTP only needs the URL. There's no authentication. Apps that can only run local servers can reach it through mcp-remote:
{
"mcpServers": {
"verified-fyi": {
"command": "npx",
"args": ["-y", "mcp-remote", "https://verified.fyi/mcp"]
}
}
}
This is the response check_website returns for stripe.com right now.
stripe.com: Trusted, trust score 100/100 (checked 27 Oct 2026) Stripe is a highly legitimate and secure financial infrastructure provider used by millions of businesses worldwide. You can trust this site with sensitive payment information and corporate data. Source: verified.fyi — https://verified.fyi/review/stripe.com
Two tools, both read-only. They return structured JSON as well as text.
The trust report for a domain or any URL on it: verdict, 0–100 score, a plain-English summary and up to five red flags, worst first. Sites we've already checked come back instantly. A new site is scanned on the spot, which usually takes under a minute. If it isn't ready after about 25 seconds, the agent is told to call again shortly.
Our published research on scam and phishing sites. Each story comes with its headline number, a short description, a link and its JSON dataset, for when you want figures to cite.
Building your own agent? GET /api/review/{domain} returns the full report, including every signal and the score for each category. No key, and CORS is open.
/llms.txt explains the site, the API and the MCP server to language models. /llms-full.txt adds the full scoring method and response fields.
Lookups of sites we've already checked are never rate limited. Brand-new scans share the website's limits, so an agent checking hundreds of unknown sites at once will be asked to wait.
When your agent checks a site for the first time, its report is published here, just like a check from the homepage. Only the site is published, never who asked. See our privacy policy.
Every result links to the full report. If your app shows our data, please name verified.fyi and keep that link.
It sums up the risk signals our checks found, and it can be wrong. Nobody can pay for a better score. Your agent gets the same verdict anyone sees on the site, with a link to the evidence behind it.
Paste any address on the homepage. You'll get the same report your agent reads.