Execution MCP Server

The Execution MCP server lets AI assistants make authenticated API calls on behalf of users. It provides one tool per API tag (~30 tools) with structured parameters, so agents see exactly what operations are available.

Server URL: https://exec.mcp.cakemail.com/api/mcp

Connect with Claude Code

claude mcp add exec-mcp \
  --transport http \
  https://exec.mcp.cakemail.com/api/mcp \
  --header "Authorization: Basic BASE64_ENCODED_CREDENTIALS"

Generate the Base64 string:

echo -n 'you@example.com:your-password' | base64

Connect with any MCP client

{
  "mcpServers": {
    "exec-mcp": {
      "type": "streamable-http",
      "url": "https://exec.mcp.cakemail.com/api/mcp",
      "headers": {
        "Authorization": "Basic BASE64_ENCODED_CREDENTIALS"
      }
    }
  }
}

Authentication

Authentication is required. Three methods are supported:

• Basic Auth — Base64-encoded email:password (recommended for MCP clients)
• Bearer token — An access token obtained from the /api/auth endpoint
• Admin API key — Set via the API_KEY environment variable on the server

How it works

Each API tag (Campaigns, Contacts, Lists, etc.) is exposed as a separate MCP tool. To use a tool, specify an operation and the relevant parameters.

Example: To list campaigns, the agent calls the campaigns tool with operation: "list".

Example: To delete a contact, the agent calls the contacts tool with operation: "delete", contact_id: "123", and confirmed: true.

Safety

• Destructive operations (DELETE) require confirmed: true. Without it, the server returns a warning instead of executing.
• Method whitelist — Only GET, POST, PATCH, DELETE methods are allowed.

Token endpoint

The server also exposes a /api/auth endpoint for obtaining access tokens:

curl -X POST https://exec.mcp.cakemail.com/api/auth \
  -H "Content-Type: application/json" \
  -d '{"username": "you@example.com", "password": "your-password"}'