About snapshot
Verifux is an MCP server that checks a website's UX and AI-agent readiness from a URL, returning a band, automated-check pass rate, key findings, and whether an agent surface is declared. This lightweight client for the Verifux audit backend helps coding agents spot issues that can make a site hard for agents to use.
Key features
- 01Run a website snapshot from a URL
- 02Return a band and check pass rate
- 03List the top findings
- 04Detect whether an agent surface is declared
Use cases
- 01Check a site's usability for AI agents
- 02Find missing structured data or page headings
- 03Verify whether an agent surface is detectable
MCP support
As an MCP server, it provides the run_snapshot tool to MCP clients and sends public URLs to the Verifux API for checks.
Configuration
Local server config is the same for every client. Copy it into your client’s MCP config file.
npx -y @verifux/mcpVERIFUX_API_URLOverride the backend the server calls. Defaults to https://api.verifux.app.Local servers need a runtime on your machine (e.g. Node.js 18+). Config file locations differ by client; see your client’s docs.
Tools
1Saved tool list · updated 1w ago
| Parameter | Type | Required | Description |
|---|---|---|---|
| url | string | Yes | The page URL to audit (any publicly reachable URL). |
Details
From the project READMEVerifux · MCP server
Let your coding agent check whether an AI agent can use a site
One tool, run_snapshot. Point it at any URL and it reports what an agent actually finds
when it arrives: a band, a pass rate, the top findings, and whether the page declares an
agent surface at all.
Zero dependencies. No sign-up. No API key.
Add it
Claude Code
claude mcp add verifux -- npx -y @verifux/mcp
Any MCP client (Claude Desktop, Cursor, and others)
{
"mcpServers": {
"verifux": {
"command": "npx",
"args": ["-y", "@verifux/mcp"]
}
}
}
Then just ask: "run a Verifux snapshot on example.com".
What comes back
A real run, pasted verbatim:
Verifux snapshot for example.com
Band: Mixed - 13 of 16 automated checks passed
Findings: 0 critical, 2 major, 1 minor
Top findings:
[MAJOR] No Organization or FAQPage structured data for AI answer engines
[MAJOR] No H1 heading on the page
[MINOR] No meta description for search engines or AI summaries to quote
AI agent surface: none detected
That last line is the one almost nobody checks. none detected means nothing an agent looks
for was found: no declared tools, no /.well-known/mcp.json, no /agents.md.
We got exactly that on our own site. The file existed. We were serving it at the wrong path, and nothing anywhere warned us. One path, and the whole surface was invisible.
Agent readiness is the only part of a site that can break without anyone noticing. No error, no alert, nothing in a dashboard. Just traffic that never arrives.
The tool
| Name | run_snapshot |
| Takes | url, any publicly reachable URL. Nothing else |
| Returns | Band (Strong / Mixed / Weak / Poor), pass rate over named automated checks, top findings, and whether an agent surface is declared |
| Env | VERIFUX_API_URL to point at a different backend. Defaults to https://api.verifux.app |
It deliberately has no score out of 100. That belongs to the full 54-checkpoint audit, and two different numbers out of 100 would contradict each other.
Prefer a terminal?
The same check as a CLI, with a CI mode that fails the build when the band drops:
npx @verifux/cli your-site.com --ci --min-band mixed
This is the free snapshot: a fast, deterministic read. The full audit lives at verifux.app.
This repo is the client shim. The audit engine behind api.verifux.app is not open source.
MIT · built by Verifux
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.