MCP Protocol Tutorial: Connect AI Agents to Real-World Tools

What Is MCP?

The Model Context Protocol (MCP) is a standard for connecting AI agents to external tools and data sources. Think of it as USB-C for AI — one protocol that connects any agent to any tool, regardless of framework.

Without MCP, every agent-tool integration is custom. Agent A uses REST to talk to GitHub, Agent B uses a different SDK for Jira, Agent C has its own Slack connector. MCP standardizes this: one protocol, one endpoint, any tool.

How MCP Works

MCP uses JSON-RPC 2.0 — a lightweight remote procedure call protocol. The agent sends a JSON request to the MCP server, the server executes the tool, and returns the result. It is stateless, language-agnostic, and easy to debug.

• Client (agent) sends a JSON-RPC request with method name and parameters
• Server validates the request, checks permissions, and executes the tool
• Server returns the result (or an error) as a JSON-RPC response
• Every call is logged with actor context, latency, and cost

Your First MCP Tool Call

import requests

# Call the MCP endpoint
response = requests.post(
    "https://dobby-ai.com/api/v1/gateway/mcp/endpoint",
    headers={
        "Authorization": "Bearer gk_user_your_key_here",
        "Content-Type": "application/json"
    },
    json={
        "jsonrpc": "2.0",
        "id": 1,
        "method": "tools/call",
        "params": {
            "name": "create_task",
            "arguments": {
                "title": "Review PR #42",
                "description": "Security review needed",
                "priority": "high"
            }
        }
    }
)

result = response.json()
print(result["result"])  # Task created successfully

Available Tool Categories

• Task Management — create, update, list, assign tasks (7 tools)
• GitHub Integration — PRs, issues, code search, branch management (7 tools)
• Memory & Knowledge — store, retrieve, search agent memories (4 tools)
• Web & Search — search the web, evaluate pages, extract data (7 tools)
• Workflow — manage multi-stage agent workflows (4 tools)
• Scheduling — create, pause, list agent schedules (4 tools)
• Approvals — get pending, approve, reject, request changes (4 tools)

Discovering Available Tools

# List all available tools
response = requests.post(
    "https://dobby-ai.com/api/v1/gateway/mcp/endpoint",
    headers={"Authorization": "Bearer gk_user_your_key_here"},
    json={
        "jsonrpc": "2.0",
        "id": 1,
        "method": "tools/list"
    }
)

tools = response.json()["result"]["tools"]
for tool in tools:
    print(f"{tool['name']}: {tool['description']}")

MCP vs REST API

Dobby supports both MCP and REST. Use MCP when your agent is an AI that needs to discover and call tools dynamically. Use REST when you are building a traditional integration with known endpoints. Both go through the same Gateway with the same authentication, policies, and audit trail.