Protocol Support¶

FastAgentic implements two complementary protocols for agentic interoperability:

Protocol	Purpose	FastAgentic Role
MCP (Model Context Protocol)	Tool and resource access	Expose agent capabilities to LLM hosts
A2A (Agent-to-Agent)	Agent collaboration	Enable agents to discover and delegate to each other

Why Two Protocols?¶

MCP answers: "What tools and resources can this agent use?" - Connects LLM applications to external capabilities - Tools, resources, and prompts exposed to AI models - Single agent ↔ host relationship

A2A answers: "What other agents can this agent collaborate with?" - Connects agents to other agents - Task delegation and multi-agent workflows - Many-to-many agent relationships

┌─────────────────────────────────────────────────────────────┐
│                      LLM Host (Claude, etc.)                │
└─────────────────────────────────────────────────────────────┘
                              │ MCP
                              ▼
┌─────────────────────────────────────────────────────────────┐
│                    FastAgentic Runtime                       │
│                                                              │
│  ┌─────────────────────────────────────────────────────┐   │
│  │                  Agent Registry (A2A)                │   │
│  │  ┌──────────┐  ┌──────────┐  ┌──────────┐          │   │
│  │  │ Agent A  │  │ Agent B  │  │ Agent C  │   ...    │   │
│  │  │(Pydantic)│  │(LangGraph│  │ (CrewAI) │          │   │
│  │  └──────────┘  └──────────┘  └──────────┘          │   │
│  └─────────────────────────────────────────────────────┘   │
│                         ▲  │  ▲                             │
│                    A2A  │  │  │  A2A                        │
│                         └──┴──┘                             │
└─────────────────────────────────────────────────────────────┘
                              │
              ┌───────────────┼───────────────┐
              ▼               ▼               ▼
         External         External       External
         A2A Agent        A2A Agent      A2A Agent

Protocol Versions¶

FastAgentic tracks the latest stable protocol specifications:

Protocol	Version	Release Date	Key Features
MCP	2025-11-25	November 2025	Tasks, Extensions, OAuth, Parallel Tools
A2A	0.3	August 2025	gRPC, Signed Cards, Extended SDK

Unified Benefits¶

By implementing both protocols, FastAgentic provides:

Single Definition, Dual Exposure
Define your agent once with decorators
Automatically exposed via MCP (for LLM hosts) and A2A (for other agents)
Internal Agent Registry
Agents deployed in FastAgentic can discover each other
A2A Agent Cards generated from adapter metadata
Task delegation without external orchestration
External Interoperability
Connect to any A2A-compliant agent (LangChain, Vertex AI, Azure AI Foundry)
Expose capabilities to any MCP-compliant host (Claude Desktop, IDEs)
Unified Authentication
Single OAuth2/OIDC configuration
MCP authorization and A2A security schemes aligned
Consistent identity across protocol boundaries

Quick Start¶

from fastagentic import App
from fastagentic.adapters.pydanticai import PydanticAIAdapter
from fastagentic.protocols import enable_a2a, enable_mcp

app = App(
    title="Support Triage",
    version="1.0.0",
)

# Enable both protocols
enable_mcp(app)      # /mcp/schema, /mcp/tools, stdio transport
enable_a2a(app)      # /.well-known/agent.json, /a2a/tasks

# Register an agent - automatically exposed via both protocols
@app.agent_endpoint(
    path="/triage",
    runnable=PydanticAIAdapter(triage_agent),
    a2a_skill="support-triage",  # A2A skill name
    mcp_tool="triage_ticket",    # MCP tool name
)
async def triage(ticket: TicketIn) -> TicketOut:
    ...

Learn More¶

MCP Implementation - Model Context Protocol details
A2A Integration - Agent-to-Agent protocol details
Templates - Starter projects for each framework

Sources¶

Protocol specifications: - MCP Specification (2025-11-25) - A2A Specification (v0.3) - A2A GitHub