Mailverdict

What the mailverdict MCP server does

The mailverdict MCP server gives MCP-compatible agents two email-intelligence tools: check_email and check_domain. The tools use the project’s maintained dataset and validation logic to identify disposable or burner domains, role-based addresses such as info@, free email providers, MX availability, and possible domain typos.

check_email accepts a complete address and can return normalized address data, the parsed user and domain, classification fields, a 0–100 score, an explanation, MX records, and a did_you_mean correction when one is available. check_domain is intended for domain-level screening. It is the privacy-conscious choice when local-part analysis is not needed, because the full email address does not leave the calling system.

How it works

The hosted mailverdict MCP server is available at https://api.mailverdict.dev/mcp using stateless Streamable HTTP and JSON-RPC 2.0. It does not require signup or an API key. A local stdio implementation is also included in the repository and exposes the same two tools.

The underlying dataset is assembled from multiple maintained public sources, deduplicated, and refreshed daily. Classification logic includes allowlist handling, parent-domain matching for subdomains, syntax checks, and MX lookups performed through Cloudflare’s DNS-over-HTTPS resolver. Email checks can skip the MX lookup with the mx=false option at the HTTP API layer; the available MCP tool parameters should be checked in the server schema before relying on that option through MCP.

Responses follow a structure intended to resemble common email-validation APIs. Signals include result and reason fields, boolean classifications, score data, and typo suggestions. The service also provides an HTTP freshness feed, but the MCP interface specifically lists the email and domain checking tools.

Setup and configuration

For the hosted service, configure an MCP client to use the POST endpoint https://api.mailverdict.dev/mcp. No credential or environment variable is required. This is suitable for clients that support remote Streamable HTTP MCP servers.

For local development from a checkout, the README provides npm run mcp, which starts the stdio transport. The project requires Node.js 20 or newer. The documented Claude Code registration runs npx tsx src/mcp.ts from the repository directory, so it is a source-checkout workflow rather than a standalone package command.

The repository also includes an SDK that calls the live API and falls back to a bundled offline snapshot after a timeout or failure. That fallback belongs to the SDK integration rather than the hosted MCP transport, so applications that need non-blocking behavior should evaluate the SDK or self-hosting options separately.

Tools and capabilities

• check_email: evaluates a complete email address for syntax, disposable status, role-account status, free-provider classification, MX information, normalization, and typo suggestions.
• check_domain: evaluates a domain without sending a complete email address to the service.
• Hosted keyless access through Streamable HTTP.
• Local stdio access for agents running the repository.
• Dataset updates performed on a daily schedule.

Limitations and notes

The classifications are advisory signals based on public lists and heuristics. False positives are possible when sources disagree or domains change, so a single boolean should not automatically block a user. The hosted service is free, rate-limited per IP, best-effort, and has no SLA. It is intended for signup and form validation, not bulk-cleaning purchased or scraped mailing lists.

The service states that it does not store email addresses per request. For MX checks, only the domain is sent to the public DNS-over-HTTPS resolver. Higher throughput, SLA, and commercial arrangements are available by contacting the provider, but optional API keys for high-volume use are listed as future work rather than a current requirement.