33.1% of the 4,640 domains of 4,707 NYSE and NASDAQ companies have no DMARC policy that stops forged email; 93.3% of the Dow 30 enforce DMARC, versus 50.5% of micro caps. See the report and infographic →
mailskop. a tool by Kinetik Bilişim

Check with AI assistants and the API

Claude, ChatGPT, Gemini, Copilot, Perplexity, Grok or DeepSeek can read the report straight from Mailskop. Type the domain, then copy the ready-made prompt or open it in your assistant. No key, no sign-up.

Ready-made prompt

Read the email security report Mailskop prepared for YOUR-DOMAIN: https://mailskop.com/en/check/YOUR-DOMAIN?format=md
Make it easy to read: start with the grade and score; show the component scores in a table (✅ full, ⚠️ partial, ❌ missing); then list the missing or misconfigured records in priority order, each with what is wrong and the exact fix. Put the findings the report marks as uncertain under a separate heading and do not present them as certain. End with a one-line verdict.

Keep the address in the prompt as it is: assistants can only open an address that appears in the chat; “use the Mailskop API” on its own is not enough.

Rather not copy it each time? Add it to Claude for good →

Add it to Claude for good (MCP connector)

Use Mailskop from Claude without copying a prompt each time.

With the connector Claude calls Mailskop itself, no address needed; you do not have to copy a prompt each time.

  1. In claude.ai open Settings → Connectors → Add custom connector.
  2. Name: Mailskop, URL: https://mailskop.com/en/mcp
  3. There is no authentication; save it. Then just say “check example.com with Mailskop” in a chat. On organization accounts an admin adds the connector.

In Claude Code: claude mcp add --transport http mailskop https://mailskop.com/en/mcp

GitHub Copilot in VS Code: MCP: Add Server → HTTP in the command palette, same URL.

API

Get JSON or Markdown from your own software; header analysis included.

No key, no sign-up, no cost. Every report lives at one address in three formats: HTML, Markdown and JSON.

GET https://mailskop.com/en/check/{domain}?format=md
Report as Markdown. For chat assistants that cannot send headers. kinetikbilisim.net
GET https://mailskop.com/en/check/{domain}?format=json
The same report as JSON. kinetikbilisim.net
GET https://mailskop.com/en/check/{domain} + Accept: text/markdown | application/json
The same address with content negotiation. Terminal tools (curl, wget) get Markdown without any header. curl https://mailskop.com/en/check/kinetikbilisim.net
GET https://mailskop.com/api/check?domain={domain}
JSON, from your own software. Disallowed for crawlers in robots.txt, so a chat assistant cannot open it.
POST https://mailskop.com/api/header
Email header analysis: the raw header in the body (plain text or {"header": "..."}). JSON report; Markdown with ?format=md, Turkish with ?lang=tr. The header is processed in the request and never stored. curl --data-binary @header.txt https://mailskop.com/api/header?format=md
POST https://mailskop.com/en/mcp
MCP server (JSON-RPC 2.0), two tools: check_domain, analyze_header

Machine descriptions: OpenAPI · llms.txt · MCP

Limit: 60 scans per minute per IP; automation calling /api/check directly gets 12 per minute.

Frequently asked

Does Mailskop have an API?

Yes, with no key, sign-up or cost. Every report is available as Markdown (?format=md) or JSON (?format=json), and the MCP server (mailskop.com/en/mcp) can be added to assistants such as Claude as a connector.

Why is “check it with the Mailskop API” not enough for an AI assistant?

A chat assistant’s web tool can only open an address that already appears in the chat; it cannot build one by filling in a domain. The ready-made prompt on this page carries the full address. The lasting fix is the MCP connector.

Can header analysis be done through the API and AI too?

Yes. Send the raw header to POST /api/header, or use the analyze_header tool of the MCP connector you added to Claude. The header is processed in the request and never stored. On the web page the header never leaves your browser.

Why does the prompt not appear in Gemini, Copilot and DeepSeek by itself?

None of the three takes a prompt passed in the address. So the button copies the prompt to your clipboard first and then opens the assistant; just paste it into the box. Grok fills in the prompt and leaves sending to you.

Can the AI relay the result wrongly?

Yes; if you see a contradiction, the measurement is right. Ask it to keep three distinctions: a missing DKIM selector does not prove DKIM is off; MX and SPF differing is not a defect on its own; a record that could not be queried is not a missing record. The ready-made prompt already says so.

Last updated: