Native MCP Server

Emails4Agents ships with a built-in Model Context Protocol (MCP) server. Wire it to Claude Code and your AI sessions get native email tools — no API calls to write, no authentication to handle.

Connect to Claude Code

Add to your ~/.claude/settings.json:

{
  "mcpServers": {
    "emails4agents": {
      "command": "npx",
      "args": [
        "-y", "mcp-remote",
        "https://api.emails4agents.com/mcp/sse",
        "--header", "X-API-Key:${E4A_API_KEY}"
      ],
      "env": {
        "E4A_API_KEY": "e4a_your_key_here"
      }
    }
  }
}

Available tools

Tool	Description
list_domains	List all configured domains
create_domain	Add a new domain with auto DNS setup
verify_domain	Check DNS propagation status
list_inboxes	List all inboxes (optionally filter by domain)
create_inbox	Provision a new inbox
send_email	Send an email from an inbox
read_messages	Read messages from an inbox
get_thread	Retrieve a full email thread
reply_to_message	Reply in-thread preserving context
search_messages	Full-text search across messages

Example Claude session

User: Create an inbox for our new outreach agent

Claude: [calls create_inbox with username="outreach", domain_id="..."]
→ Created [email protected] (inbox-uuid: abc123)

User: Send a cold email to the list in leads.csv

Claude: [reads leads.csv, calls send_email for each row]
→ Sent 47 emails from [email protected]

User: Did anyone reply?

Claude: [calls read_messages, filters direction=inbound]
→ 3 replies received. 2 interested, 1 unsubscribe.

Why MCP matters for email

Email has always been awkward for AI agents — you’d need to either scrape a webmail UI or implement IMAP in your agent harness. MCP turns email into a first-class tool call. Claude treats send_email the same way it treats read_file or bash — just another tool in the session.