What is MCP and Why It Matters for AI Agents

Model Context Protocol (MCP) is an open standard that lets AI assistants connect to external tools and data sources. Think of it as a universal adapter that lets AI agents interact with the real world — databases, APIs, file systems, and yes, email.

The Problem MCP Solves

Before MCP, giving an AI assistant access to external tools required:

• Custom integrations for each AI provider
• Complex authentication flows
• Inconsistent interfaces across tools
• Security concerns with direct API access

MCP standardizes this. One protocol, universal compatibility.

How MCP Works

MCP follows a client-server model:

1. MCP Server — A service that exposes tools (like Docmailer)
2. MCP Client — The AI assistant (Claude Code, Cursor, etc.)
3. Protocol — JSON-RPC over stdio, HTTP, or SSE

Example Flow

User: "Check my test inbox for new emails"

Claude → MCP Client: tool_call("list_emails", {inbox_id: "abc"})
MCP Client → Docmailer Server: JSON-RPC request
Docmailer Server → Docmailer API: GET /inbox/abc/emails
Docmailer API → Docmailer Server: {emails: [...]}
Docmailer Server → MCP Client: {result: {emails: [...]}}
MCP Client → Claude: "You have 3 new emails..."

Claude: "You have 3 new emails. The most recent is 
         'Verify your account' from noreply@app.com"

MCP vs Direct API Access

Why not just give the AI your API key?

• Security: MCP servers can implement fine-grained permissions
• Consistency: Same interface across all tools
• Discovery: AI can ask "what tools do you have?"
• Streaming: Real-time updates via SSE

Docmailer's MCP Implementation

Docmailer provides an MCP server that exposes these tools:

{
  "tools": [
    {
      "name": "list_inboxes",
      "description": "List all email inboxes"
    },
    {
      "name": "create_inbox", 
      "description": "Create a new test inbox"
    },
    {
      "name": "list_emails",
      "description": "List emails in an inbox"
    },
    {
      "name": "read_email",
      "description": "Read full email content"
    },
    {
      "name": "send_email",
      "description": "Send an email (Premium)"
    }
  ]
}

Setting Up MCP

Claude Code

claude mcp add --transport sse docmailer "https://docmailer.org/mcp/sse?token=YOUR_TOKEN"

Cursor / Other Clients

Add to your MCP config file:

{
  "servers": {
    "docmailer": {
      "url": "https://docmailer.org/mcp/sse",
      "auth": {
        "type": "bearer",
        "token": "YOUR_TOKEN"
      }
    }
  }
}

The Future of AI Tools

MCP is still early, but adoption is growing fast. Major players are building MCP servers:

• Databases (Postgres, SQLite)
• Browser automation (Playwright)
• File systems and Git
• Communication tools (Slack, Email)

As more services add MCP support, AI agents become genuinely capable of handling complex, multi-step tasks autonomously.