Email validation MCP server for Claude and other AI assistants
Connect Mail7 to your assistant through the Model Context Protocol and it can verify an address, clean a list of signups or explain why your mail lands in spam, without you opening another tab.
Coming soon on npm as mail7-mcp. The source is open now and runs locally.
Four tools your assistant can call
The server hands the model a small, sharp toolkit. It decides when to reach for each one:
validate_email
One address: syntax, MX records, disposable-domain database and a live SMTP mailbox probe.
validate_emails
Up to 50 addresses per call, returned with a per-address verdict and a summary count.
check_domain
A graded audit of MX, SPF, DKIM, DMARC, MTA-STS, TLS-RPT, DNSSEC and BIMI, with a fix for each problem.
check_spf
The SPF record itself: mechanisms, lookup count and the misconfigurations that break alignment.
An assistant that will not throw away good addresses
A language model asked to clean a list tends to treat anything uncertain as junk. Mail7 answers in three states, and every tool description says in plain words what the middle one means, so the model keeps what it could not confirm:
Mailbox confirmed. Safe to send.
Confirmed invalid. Remove.
Could not be checked. Keep, do not reject.
What you can ask for
Once the server is connected, the work happens in ordinary sentences:
- Is [email protected] a real mailbox?
- Take the signups in this CSV and tell me which ones will bounce.
- Our newsletter keeps landing in spam - check what is wrong with our domain.
- Before I import these leads into the CRM, drop the confirmed-invalid ones.
Connect it to your assistant
-
1
Get an API key
Create one in your Mail7 account. The server also runs without a key at anonymous limits.
-
2
Add the server to your client
In Claude Code that is one command: claude mcp add mail7 --env MAIL7_API_KEY=your_key -- npx -y mail7-mcp. Claude Desktop, Cursor and Windsurf take the same command in their JSON config.
-
3
Ask in plain language
The assistant picks the right tool by itself. Node 18 or newer is the only requirement.
Free single checks, paid plans for volume
Single checks are free and need no account. Paid plans raise your rate limit and unlock full list cleaning, and you are only charged for definite Valid and Not Valid results. Unknown checks are always free.
Related integrations
Frequently asked questions
What is an MCP server?
Model Context Protocol is an open standard for connecting AI assistants to external tools. An MCP server exposes a set of tools that the assistant can call on its own, so Claude or Cursor can validate an email without you leaving the chat.
Which AI assistants does it work with?
Any MCP client. That includes Claude Desktop, Claude Code, Cursor, Windsurf, Cline and Zed.
Do I need an API key?
The server runs without one, but anonymous callers are limited to 5 checks per minute and a 25-address bulk sample. An API key raises both and is what you want for real list cleaning.
Will the assistant delete addresses it could not verify?
No. Every tool description states that Unknown means unverifiable, not invalid, and instructs the model to keep those addresses and present them separately.
How many addresses can it check at once?
Up to 50 per call. Each address is a live SMTP conversation, so the assistant works through a longer list in chunks rather than in one request.
Are Unknown results billed?
No. You are charged only for definite Valid and Not Valid results; Unknown checks are always free.