Building Multi-Protocol Systems — A2A Protocol & Google ADK

[ ABORT TO HUD ]
SEQ. 1
SEQ. 2
SEQ. 3
SEQ. 4
SEQ. 5

Building Multi-Protocol Systems

🌐 A2A Protocol & Google ADK⏱ 12 min⭐ 120 BASE XP⌨ HANDS-ON LAB

MCP + A2A + ADK: The Full Stack

Production agent systems in 2026 use multiple protocols together. Here's how they fit:

The Three-Protocol Architecture

┌─────────────────────────────────────────────────┐
│              YOUR AGENT (built with ADK)         │
│                                                  │
│  ┌────────────────┐     ┌─────────────────────┐ │
│  │  MCP Clients   │     │   A2A Client         │ │
│  │  (Tools/Data)  │     │   (Agent Delegation) │ │
│  └───────┬────────┘     └──────────┬──────────┘ │
└──────────┼─────────────────────────┼────────────┘
           │                         │
    ┌──────▼──────┐          ┌──────▼──────────┐
    │ MCP Servers │          │ Remote A2A      │
    │ • Database  │          │ Agents          │
    │ • GitHub    │          │ • Travel Agent  │
    │ • Slack     │          │ • Legal Agent   │
    │ • Files     │          │ • Finance Agent │
    └─────────────┘          └─────────────────┘

When to Use Which

NeedProtocolExample
Read a databaseMCPQuery customer records via MCP server
Call an APIMCPSend a Slack message via MCP tool
Delegate a complex taskA2AAsk a travel agent to book a trip
Get a second opinionA2AAsk a legal agent to review a contract
Orchestrate everythingADKBuild the central agent with sub-agents

Production Implementation

from google.adk.agents import Agent
from google.adk.agents.remote_a2a_agent import RemoteA2aAgent
from google.adk.tools.mcp_tool import MCPToolset
from google.adk.tools.mcp_tool.mcp_session_manager import StreamableHTTPConnectionParams

# MCP tools for direct data access
db_mcp = MCPToolset(
    connection_params=StreamableHTTPConnectionParams(url="https://mcp.internal/postgres"),
    tool_filter=["query_customers"],
)
slack_mcp = MCPToolset(
    connection_params=StreamableHTTPConnectionParams(url="https://mcp.internal/slack"),
    tool_filter=["send_message"],
)

# A2A clients for agent delegation
travel_agent = RemoteA2aAgent(name="travel_agent", agent_card="https://travel.example.com/.well-known/agent-card.json")
legal_agent = RemoteA2aAgent(name="legal_agent", agent_card="https://legal.example.com/.well-known/agent-card.json")

# Build the orchestrator
orchestrator = Agent(
    name="executive_assistant",
    model="gemini-4-argon",
    tools=[db_mcp, slack_mcp],
    sub_agents=[travel_agent, legal_agent],
    instruction="""You are an executive assistant. 
    Use MCP tools for data access (database, Slack).
    Delegate to the travel agent for booking tasks.
    Delegate to the legal agent for contract review."""
)

Integration Checklist

  • ☐ Publish your Agent Card at /.well-known/agent-card.json
  • ☐ Register MCP servers for all data/tool access
  • ☐ Discover and validate remote A2A agents before production
  • ☐ Implement input-required state handling for A2A tasks
  • ☐ Set up OpenTelemetry for cross-protocol observability
  • ☐ Rate-limit A2A calls to prevent cascade failures
  • ☐ Authenticate all inter-agent communication (OAuth 2.0)
🌐 The Big Picture: MCP is the agent's hands (tools). A2A is the agent's network (colleagues). ADK is the agent's skeleton (structure). Together, they create agents that can do anything a human knowledge worker can do.
⌨ HANDS-ON LABBridge A2A Agents to MCP Server Tools
⭐ +150 XP

Deploy an agent gateway that translates A2A JSON-RPC delegation tasks into downstream Model Context Protocol (MCP) tool calls.

1Test protocol bridging between A2A and MCP.
lab-sandbox — simulated environment
INFINITY LAB SANDBOX v2.6 — simulated shell
Type the command for the current objective. Helpers: "hint", "solution", "clear".
$
OBJECTIVE 1 / 1 — type "hint" if stuck
SYNAPSE VERIFICATION
QUERY 1 // 3
When should you use A2A instead of MCP?
For reading files
When you need to delegate a complex task to a specialized agent, not just call a tool
For database queries
For authentication