Skip to content

Vetto MCP

Every signup form accepts an address it cannot deliver to, and the bill arrives a month later as an unpaid trial. The address is the cheapest thing a person types and the most expensive thing to get wrong: a typo locks out a real buyer, a throwaway domain fills the trials, and one Gmail account becomes seven registrations because dots and plus tags are free. Vetto turns that judgement into four read-only tools on a public endpoint with no account and no key.

Server type:  remote Streamable HTTP, keyless
Endpoint:     https://usevetto.com/mcp
Server:       vetto 1.0.0 (serverInfo), title "Vetto: fake signup checks", protocol 2025-03-26
Capabilities: tools only (no resources, no prompts)
Tools:        4 measured, all annotated readOnlyHint: true
Registry:     no record found under this name (the search returns si.vettoz/vettoz, a different product)
Probe:        initialize, tools/list and six tools/call requests in one session

What the server answered

initialize returned 200 with no credential and an instruction block that states the product's four questions in its own words: disposable and throwaway domains, domains that cannot receive mail, typos such as gmial.com, privacy relays (allowed, because they reach a real inbox), and Gmail aliases of the same inbox. tools/list returned four:

check_email     one address, judged the way a signup form judges it
check_domains   up to 50 domains at once, for the domains behind recent signups
same_inbox      two addresses compared after Gmail normalisation
how_to_block    setup steps for a given stack

Driven: a typo, a throwaway and an alias

Six calls were made, and the first two show why the verdict is more useful than a boolean.

check_email on jdoe@gmail.co returned BLOCK, reasons: ["typo", "dead"], suggestion: "jdoe@gmail.com", domainStatus: "null_mx" and an empty mx array. The suggestion is the part that saves the customer: an assistant can answer did you mean jdoe@gmail.com inside the signup conversation rather than showing a validation error.

check_email on billing@mailinator.com returned BLOCK with reasons: ["role", "disposable"] and domainStatus: "skipped". check_email on team+1@googlemail.com returned ALLOW with reasons: ["freemail", "role"].

Read together, those two answers define the rule precisely: a role address alone does not block. A role address on a freemail domain is allowed with the tag recorded; the same role shape on a disposable domain is blocked because the disposable signal carries the verdict. A tool that blocked on role as a flag would have rejected a real team inbox, and one that ignored it would have let the throwaway through with no note. The response keeps both signals and ranks them.

same_inbox on john.doe+promo@gmail.com and johndoe@googlemail.com returned same: true with both normalising to johndoe@gmail.com; the control pair john@gmail.com and jane@gmail.com returned same: false. That is the trial-abuser check stated as arithmetic rather than as a claim.

check_domains on four domains returned a summary of checked: 4, flagged: 2: gmail.com ok, mailinator.com disposable, gmial.com disposable, acme-supply.com ok. Note that gmial.com is flagged as a disposable domain, not as a typo, and the vendor's own page makes the distinction worth knowing: gmai.com carries a live mail server, so an MX-only check passes it, while gmial.com resolves to something the vendor already classifies. Two different lookalike domains, two different signals.

how_to_block on next.js returned a server-side recipe rather than a marketing page: POST to https://usevetto.com/v1/audit with the domain list, reject when results[domain].disposable is true or the status is nxdomain, null_mx or no_mail, and fail open on timeout. That last clause is the operational decision most form guards get wrong in the other direction, and it is stated in the tool output.

What the vendor states, and what it does not

From the vendor site read on 2026-11-12: one script tag (/v.js with a site key) on the signup form, free on one site with no card, free for Next.js, Supabase, Webflow, Framer and WordPress, failing open after 800 ms, and eight signals folded into one verdict per signup. The headline claim is that Stripe saw 6.2x more abusive free trials between November 2025 and February 2026, sourced to Stripe, alongside a ~10 percent figure for registrants who typed their email wrong, sourced to Discourse Meta. There is a Founding offer of Pro for life at 69 USD once for the first 100 buyers until November 27.

Two honest limits. The server's own instruction block says nothing sent to these tools is stored; that is the vendor's statement about its own endpoint, quoted as such and not a claim about an account you connect. And no official MCP registry record was found under this product name, so the endpoint evidence above is the whole of the technical record: keyless, four tools, six driven calls.

Where it sits

This is the first signup-integrity entry in this catalog. Vibe Shield MCP scans the code that accepts the form; this judges the value typed into it. Emailchaser MCP owns the outbound side, where a bad address costs a send rather than a trial.

Does Vetto need an API key to answer a check?

No. initialize and tools/list answered 200 with no credential, and all six tools/call requests returned verdicts. Whether an account changes rate limits or coverage is not published on the endpoint and was not measured here.

Can it tell two Gmail addresses apart from the same person?

Not apart - together. same_inbox normalises dots, +tags and googlemail.com, so john.doe+promo@gmail.com and johndoe@googlemail.com are reported as one inbox, which is how a seventh trial from the same person is detected.

What happens when the check times out?

The vendor's own recipe tells the caller to fail open, so a slow check lets a real customer through rather than blocking a signup. That is a product decision to verify against your own form before switching the script on.

Integration with CorpusIQ

CorpusIQ is the connector layer that lets AI assistants read the systems a business already runs on - accounting, payments, CRM, email, files and analytics - and answer with the source named. It is read-only by design: it retrieves data and never writes to, or acts in, a connected system, which is what makes it safe to point at live accounts.

This page is part of the CorpusIQ directory, the reference we maintain for the MCP server and agent-skill ecosystem. Run alongside CorpusIQ, a third-party server like this one can be combined with a company's own business data in a single answer.

See what CorpusIQ connects

AI answers on your data. See where every number comes from.

Connect your tools and ask in ChatGPT, Claude, or Perplexity. Enterprise-grade answers in minutes. From $29.95/month.

Start free, no credit card
Contact Terms Privacy Pricing