Agenticx A2a Connector

AgenticX A2A Connector

Guide for building distributed, inter-communicating agent systems using the A2A protocol.

What is A2A?

A2A (Agent-to-Agent) is a protocol that enables agents to discover each other's capabilities and invoke them as if they were local tools. This allows building distributed agent systems where specialized agents collaborate across network boundaries.

Core Components

Agent Cards

An AgentCard declares what an agent can do:

from agenticx.protocols import AgentCard, Skill

card = AgentCard(
    name="Research Agent",
    description="Specializes in web research and report generation",
    url="http://research-agent:8000",
    skills=[
        Skill(
            name="web_research",
            description="Search the web and compile findings",
            parameters_schema={
                "type": "object",
                "properties": {
                    "query": {"type": "string"},
                    "depth": {"type": "integer", "default": 3}
                },
                "required": ["query"]
            }
        ),
        Skill(
            name="generate_report",
            description="Generate a structured report from research data",
            parameters_schema={
                "type": "object",
                "properties": {
                    "topic": {"type": "string"},
                    "format": {"type": "string", "enum": ["markdown", "html"]}
                },
                "required": ["topic"]
            }
        )
    ]
)

Using Remote Agent Skills as Tools

Single Skill

from agenticx.protocols import A2ASkillTool

tool = A2ASkillTool(
    agent_url="http://research-agent:8000",
    skill_name="web_research"
)

# Use like any local tool
result = tool.run(query="latest AI trends", depth=5)

All Skills from an Agent

from agenticx.protocols import A2ASkillToolFactory

factory = A2ASkillToolFactory()
tools = factory.create_tools(agent_card=card)

# tools is a list of A2ASkillTool instances, one per skill
for t in tools:
    print(f"Tool: {t.name} — {t.description}")

A2A Client

Low-level client for direct communication:

from agenticx.protocols import A2AClient

client = A2AClient(base_url="http://research-agent:8000")

# Discover capabilities
card = client.get_agent_card()

# Invoke a skill
result = client.invoke_skill(
    skill_name="web_research",
    parameters={"query": "quantum computing", "depth": 3}
)

Building an A2A-Enabled Agent

Expose your agent as an A2A service:

from agenticx import Agent
from agenticx.protocols import A2AServer

agent = Agent(
    id="specialist",
    name="Data Specialist",
    role="Data Analysis",
    goal="Analyze datasets",
    organization_id="team-a"
)

server = A2AServer(agent=agent, port=8001)
server.register_skill(
    name="analyze_data",
    handler=my_analysis_function,
    description="Analyze a dataset and return insights"
)
server.start()

Multi-Agent Architecture Pattern

┌─────────────────┐     A2A      ┌──────────────────┐
│  Orchestrator    │────────────→│  Research Agent   │
│  Agent           │             │  (port 8001)      │
│  (port 8000)     │             └──────────────────┘
│                  │     A2A      ┌──────────────────┐
│                  │────────────→│  Analysis Agent   │
│                  │             │  (port 8002)      │
└─────────────────┘             └──────────────────┘

The Orchestrator discovers remote agents via their AgentCards and invokes their skills as tools within its own workflow.

CLI for A2A

# Start an agent as an A2A service
agx serve --port 8001

# The serve command exposes:
# - GET  /.well-known/agent-card  → AgentCard JSON
# - POST /tasks/submit            → Invoke skills

Best Practices

1. Version your skills — include version in AgentCard metadata
2. Schema validation — always define parameters_schema for skills
3. Timeout handling — set reasonable timeouts for remote calls
4. Retry logic — implement retries for network failures
5. Health checks — use /health endpoints before routing traffic
6. Security — authenticate A2A calls in production environments