MCP Integration

Extend your agents with external tools and services using the Model Context Protocol.

Stable

This page covers connecting MCP servers to your agents. For protocol details and building your own MCP server, see MCP Protocol.

Overview

MCP (Model Context Protocol) allows agents to:

Access external tools: Search, calculations, APIs
Use resources: Files, databases, services
Execute prompts: Pre-defined interaction templates

Quick Start

Dashboard

Browse and connect MCP servers at Dashboard → MCPs.

API


curl -X POST "https://cloud.milady.ai/api/v1/mcps" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Weather MCP",
    "url": "https://mcp.eliza.ai/weather",
    "agentId": "agent_abc123"
  }'

MCP Registry

elizaOS Cloud provides a curated registry of MCP servers:

Server	Description	Tools
Weather	Current weather and forecasts	`get_weather`, `get_forecast`
Search	Web search capabilities	`search_web`, `search_images`
Crypto	Cryptocurrency data	`get_price`, `get_market_data`
Calculator	Math operations	`calculate`, `convert_units`
Time	Time and timezone tools	`get_time`, `convert_timezone`

Browse Registry


curl -X GET "https://cloud.milady.ai/api/mcp/registry" \
  -H "Authorization: Bearer YOUR_API_KEY"


{
  "servers": [
    {
      "id": "weather",
      "name": "Weather MCP",
      "description": "Get weather data for any location",
      "url": "https://mcp.eliza.ai/weather",
      "tools": [
        {
          "name": "get_weather",
          "description": "Get current weather for a location",
          "inputSchema": {
            "type": "object",
            "properties": {
              "location": { "type": "string" }
            }
          }
        }
      ]
    }
  ]
}

Connecting MCP Servers

Select a Server

Browse the MCP registry or use a custom server URL.

Configure Connection

Provide the server URL and any required authentication.

Link to Agent

Connect the MCP server to your agent.

Test

Verify the tools are available in the agent.

Connect via API


curl -X POST "https://cloud.milady.ai/api/v1/mcps" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "My MCP Server",
    "url": "https://my-mcp-server.com",
    "agentId": "agent_abc123",
    "authentication": {
      "type": "bearer",
      "token": "mcp_token_xxx"
    }
  }'

Using MCP Tools

When MCP servers are connected, agents can use their tools automatically:


const response = await fetch("https://cloud.milady.ai/api/v1/chat/completions", {
  method: "POST",
  headers: {
    Authorization: "Bearer YOUR_API_KEY",
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    model: "agent_abc123", // Agent with MCP connected
    messages: [{ role: "user", content: "What's the weather in Tokyo?" }],
  }),
});

The agent will automatically invoke the get_weather tool and include the result in its response.

MCP Protocol

JSON-RPC Interface

elizaOS Cloud exposes an MCP-compatible JSON-RPC interface:


curl -X POST "https://cloud.milady.ai/api/mcp" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "jsonrpc": "2.0",
    "method": "tools/list",
    "id": 1
  }'

Available Methods

Method	Description
`tools/list`	List available tools
`tools/call`	Execute a tool
`resources/list`	List available resources
`resources/read`	Read a resource
`prompts/list`	List available prompts
`prompts/get`	Get a prompt

Execute Tool


{
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "name": "get_weather",
    "arguments": {
      "location": "San Francisco"
    }
  },
  "id": 2
}

Response:


{
  "jsonrpc": "2.0",
  "result": {
    "content": [
      {
        "type": "text",
        "text": "Current weather in San Francisco: 65°F, Partly Cloudy"
      }
    ]
  },
  "id": 2
}

Custom MCP Servers

Server Requirements

Your MCP server must implement:

HTTP Transport: Accept POST requests
JSON-RPC 2.0: Standard JSON-RPC protocol
MCP Methods: initialize, tools/list, tools/call

Basic Server Example


// Simple MCP server with Express
const express = require("express");
const app = express();
 
app.use(express.json());
 
const tools = [
  {
    name: "greet",
    description: "Greet a person",
    inputSchema: {
      type: "object",
      properties: {
        name: { type: "string" },
      },
      required: ["name"],
    },
  },
];
 
app.post("/mcp", (req, res) => {
  const { method, params, id } = req.body;
 
  if (method === "tools/list") {
    return res.json({
      jsonrpc: "2.0",
      result: { tools },
      id,
    });
  }
 
  if (method === "tools/call") {
    const { name, arguments: args } = params;
    if (name === "greet") {
      return res.json({
        jsonrpc: "2.0",
        result: {
          content: [{ type: "text", text: `Hello, ${args.name}!` }],
        },
        id,
      });
    }
  }
 
  res.json({
    jsonrpc: "2.0",
    error: { code: -32601, message: "Method not found" },
    id,
  });
});
 
app.listen(3000);

Demo Servers

elizaOS Cloud provides demo MCP servers for testing:

Server	URL	Description
Weather	`/api/mcps/weather`	Weather data demo
Time	`/api/mcps/time`	Time and timezone demo
Crypto	`/api/mcps/crypto`	Cryptocurrency data demo

Test Demo Server


curl -X POST "https://cloud.milady.ai/api/mcps/weather" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "jsonrpc": "2.0",
    "method": "tools/call",
    "params": {
      "name": "get_weather",
      "arguments": { "location": "New York" }
    },
    "id": 1
  }'

Managing MCP Connections

List Connected MCPs


curl -X GET "https://cloud.milady.ai/api/v1/mcps?agentId=agent_abc123" \
  -H "Authorization: Bearer YOUR_API_KEY"

Disconnect MCP


curl -X DELETE "https://cloud.milady.ai/api/v1/mcps/mcp_abc123" \
  -H "Authorization: Bearer YOUR_API_KEY"

Publish Custom MCP

Share your MCP server in the registry:


curl -X POST "https://cloud.milady.ai/api/v1/mcps/{mcpId}/publish" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "visibility": "public",
    "description": "My custom MCP server",
    "category": "utilities"
  }'

Best Practices

Minimal Tools — Only connect tools your agent actually needs to reduce latency
Test Thoroughly — Verify tools work correctly before deploying to production
Handle Errors — MCP calls can fail; implement graceful fallbacks in your agent
Monitor Usage — Track tool invocations in Dashboard for optimization

Next Steps

Create agents with MCP tools

AI Agents

Deep dive into MCP protocol

MCP Protocol

Agent-to-Agent communication

A2A Protocol