Check business emails from your AI assistant with MCP

The isBusinessEmail MCP server lets an AI assistant such as Claude, ChatGPT, Cursor or VS Code check email addresses for you: is jane@acme.io a work email, which of this week’s sign-ups are personal or disposable, does Acme run Google Workspace or Microsoft 365. Add https://api.isbusinessemail.com/mcp as a connector, sign in once, and ask in plain language. It’s free, and every check counts toward the same daily quota as the API.

What MCP is

The Model Context Protocol (MCP) is an open standard for connecting AI assistants to outside tools and data. A service publishes an MCP server that describes its tools, and any MCP-capable client can call them in the middle of a conversation.

For email classification, that means the assistant stops guessing. Ask a model whether acme.io uses Microsoft 365 and it can only reason from what it remembers. Connect the isBusinessEmail server and the assistant gets the domain’s real MX records, mail provider and Microsoft tenant from the same engine that answers API calls.

What the server does

Server URL	https://api.isbusinessemail.com/mcp
Transport	Streamable HTTP (JSON responses, no sessions)
Sign-in	OAuth 2.1 with PKCE through your isBusinessEmail account, or an API key as a Bearer token
Cost	Free. Each check counts toward your daily quota, like an API call

It offers five tools:

Tool	What it does	Counts as
check_email	Classifies one address: category, is_business, recommendation, reasons, typo suggestion, mail provider, Google Workspace / Microsoft 365	1 check
check_domain	The same for a domain, for when you only care about the company	1 check
check_batch	Up to 100 addresses or domains in one call, results in input order	1 check per item
get_usage	Tier, limits, checks used and remaining today, last 30 days	free
report_wrong_verdict	Reports a misclassified address or domain; a person reviews it	free (50 a day)

The three check tools take an optional policy: b2b (the default), strict or lenient. The policy changes the recommendation, not the category; categories and policies has the full mapping.

Connect your assistant

Every client needs the same URL. The MCP page has copy-ready snippets, and the MCP docs have step-by-step instructions and troubleshooting.

Claude (web, desktop and mobile)

1. Open Settings → Connectors and choose Add custom connector.
2. Name it isBusinessEmail and paste https://api.isbusinessemail.com/mcp as the URL.
3. Select Connect, sign in to isBusinessEmail (or create a free account) and select Allow access.

On Team and Enterprise plans, an owner adds the connector for the organization first.

Claude Code

claude mcp add --transport http isbusinessemail https://api.isbusinessemail.com/mcp

Then run /mcp inside Claude Code and choose Authenticate. A browser window opens for sign-in.

ChatGPT

Add a custom connector (in developer mode, depending on your plan) with the server URL and OAuth authentication. ChatGPT registers itself and opens the sign-in page.

Cursor and VS Code

Cursor reads ~/.cursor/mcp.json (or .cursor/mcp.json in a project) and shows Needs login next to the server; select it to sign in. VS Code reads .vscode/mcp.json:

{
  "servers": {
    "isbusinessemail": { "type": "http", "url": "https://api.isbusinessemail.com/mcp" }
  }
}

Clients without OAuth

Send a secret API key from your dashboard as a header instead:

{
  "mcpServers": {
    "isbusinessemail": {
      "url": "https://api.isbusinessemail.com/mcp",
      "headers": { "Authorization": "Bearer ibe_live_your_key" }
    }
  }
}

Keep that file out of shared repositories. OAuth is the safer option: when you select Allow access, the app receives its own secret key, named MCP · <app name>. It shows up in your dashboard with its own usage, revoking it disconnects only that app, and connecting the same app again replaces its previous key.

Things to ask

You don’t need tool names. Ask the way you’d ask a colleague, and the assistant picks the tool.

“Is acme.io a business domain, and do they use Google Workspace?”

The assistant calls check_domain once. The parts of the answer it reads look like this:

{
  "domain": "acme.io",
  "category": "business",
  "is_business": true,
  "recommendation": "allow",
  "mail": { "has_mx": true, "provider": "google_workspace", "dmarc_policy": "reject" },
  "workspace": {
    "google_workspace": { "detected": true, "evidence": ["mx_google", "dkim_google"] },
    "microsoft_365": { "detected": false, "tenant_id": null, "auth": null },
    "other_suite": null,
    "pending": false
  }
}

Google Workspace and Microsoft 365 can both be detected for one domain, for example when a company signs in with Entra ID but keeps its mail on Google, or is in the middle of a migration. If pending is true, detection is still running in the background; ask again a little later. Workspace detection explains the evidence, and which email provider does a company use shows how to read the same DNS records yourself.

“Which of these 40 leads are on Microsoft 365?”

Paste the list and the assistant can send it to check_batch in one call: 40 items, 40 checks, results in the order you gave them. It can then filter on workspace.microsoft_365.detected and, where it’s available, show whether the tenant signs in with passwords managed in Entra ID or federates to another identity provider such as Okta.

If several leads work at the same company, ask it to check each company domain once instead of every address. You get the same Workspace and Microsoft 365 answer for fewer checks. This is the question behind sales routing by workspace, and lead scoring by email domain covers what to do with the answer.

“Clean up this CSV of sign-ups”

Paste the rows, or attach the file if your client can read files, and be specific about the output:

Check every address in this list. Give me a table with email, category and recommendation, and include any did_you_mean suggestion. List disposable and relay addresses separately.

check_batch takes up to 100 items per call, so the assistant works through a longer list in batches. Watch the arithmetic: 500 rows are 500 checks, which is more than a new account’s 100 a day. Remove duplicates before you paste, and for a large file you’ll process more than once, use the bulk email checker or call the batch endpoint from a script. The CRM hygiene playbook covers what to do with the categories once you have them.

Follow-ups that work well

• “Why was contoso.com classified as business? Show me the reasons.”
• “Run that again with the strict policy.”
• “How many checks do I have left today?” (get_usage, free)
• “acme-mail.com is our customer’s real company domain; report that it was misclassified.” (report_wrong_verdict, free, and a person reviews it before any list changes)

What the assistant gets back

Every tool returns two things: a short text summary for the assistant, and the full API response as structured content, with the same fields as POST /v1/check. The summary is what the model usually repeats to you; the structured content holds every field it can filter, sort or quote.

Three habits keep the answers trustworthy:

1. Ask for the fields when you’ll act on the result. “Show category, recommendation and reasons for each” makes the assistant quote the response instead of paraphrasing it.
2. Remember what the check is. It classifies the address from its domain, DNS and curated lists. It doesn’t confirm that a mailbox exists, and it never emails or probes the address. Email verification vs classification explains the difference.
3. Keep decisions that run on their own in code. An assistant is the right tool for questions and one-off lists. A sign-up form or a CRM workflow should call the API directly, with a timeout and a fail-open path, as in checking for a business email in code.

Quotas

MCP checks use your account’s rate limits and daily quota, exactly like API calls, and show up in your usage the same way.

• New accounts get 100 checks a day. Accounts move up to 1,000 a day automatically after 7 clean days, and higher limits are available on request.
• check_email and check_domain count 1 check, check_batch counts 1 per item, and get_usage and report_wrong_verdict are free.
• Quotas reset at 00:00 UTC. Once the quota is used up, tools return a quota_exceeded error until then.

Privacy

The MCP server follows the same rules as the API. It only sees the addresses and domains your assistant sends, never emails or probes them, and stores domains rather than full addresses (local parts only as a keyed hash). Privacy and data has the full table.

There is one more party in the loop, though: your AI provider. Whatever you paste into the conversation goes to the assistant before any tool runs, under your agreement with that provider and your plan’s data settings. So:

• Don’t paste data you aren’t allowed to share with your AI provider. If your company has rules about customer data in AI tools, they apply here.
• Send less. Delete names, phone numbers and other columns before you paste a CSV, and send domains instead of addresses when you only need the company.
• Keep tool approval on. The assistant decides when to call a tool, and most clients ask before running one. Check your client’s settings.

Get started

1. Create a free account, or let the sign-in screen create one when you connect.
2. Add https://api.isbusinessemail.com/mcp to your assistant, using the steps on the MCP page.
3. Ask “How many checks do I have left today?” It’s free, and it confirms the connection works.
4. Try a real question: one domain, then a short list.

If a client misbehaves, the troubleshooting table in the MCP docs covers expired sign-ins, revoked keys and the active-key limit.