Agent2Agent Protocol

Active
GitHub Shell Apache-2.0

Description

A2A (Agent-to-Agent) Protocol is an open protocol by Google enabling interoperability and collaborative communication between AI agents built across different frameworks and vendors.

Key Features

  • Standardized agent communication — JSON-RPC 2.0 over HTTP(S) protocol
  • Agent discovery mechanism — Agent Cards describe capabilities, interaction modalities and connection info
  • Flexible interaction modes — supports synchronous request/response, SSE streaming and async push notifications
  • Rich data exchange — handles text, files and structured JSON data
  • Multi-language SDKs — official SDKs in Python, Go, JS, Java, .NET and Rust
  • Enterprise-grade security — built-in authentication, authorization and observability support

Use Cases

💡 Enable AI agents built on different frameworks to discover and collaborate on complex tasks
💡 Build cross-organizational multi-agent workflows combining specialized agent capabilities
💡 Add A2A-compatible interfaces to existing agent applications for external agent invocation
💡 Connect different vendor agent platforms within enterprises using A2A
💡 Build agent marketplaces where agents discover and invoke each other's services via standard protocols

Strengths & Limitations

Strengths

  • Actively maintained, recent updates
  • High community interest (25.4k stars)
  • Permissive open-source license (Apache-2.0)
  • Responsive to issues, low backlog

Quick Start

pip install a2a-sdk

# Create an A2A Server
from a2a.server import A2AServer
from a2a.types import AgentCard

server = A2AServer(
    agent_card=AgentCard(
        name="My Agent",
        description="A sample agent",
        url="http://localhost:8000",
        skills=[{"id": "greet", "name": "Greeting"}]
    )
)
server.run()

Related Projects

GPTeam

1.7k · Python
Active A

An open-source multi-agent simulation platform where multiple LLM-powered agents collaborate on complex tasks in shared environments, with customizable roles, memory systems, and environment interaction for studying multi-agent collaboration.

pythonmulti-agentagent +3
  • · GPT-4 powered multi-agent simulation with autonomous collaboration on predefined goals
  • · Agent memory and reflection system inspired by Generative Agents research paper
  • · Spatial movement: agents move between locations and perform tasks in different environments

Cultivation World Simulator

2.0k · Python
Active A

An open-source Cultivation World Simulator using Agentic Workflow to create a dynamic, emerging Xianxia world. Showcases multi-agent collaboration in complex scenarios.

multi-agentworkflowagent +1
  • · Every NPC is an independent LLM-driven agent with unique personality, memory, relationships, and decision-making logic
  • · Emergent storytelling with no preset scripts — sect wars, rivalries, and character arcs arise from world rules and AI interactions
  • · Rich cultivation world system including spirit roots, realms, techniques, sects, alchemy, artifacts, and martial tournaments

BettaFish

42.0k · Python
Active A+

An accessible multi-agent sentiment analysis assistant that breaks filter bubbles, reveals true public opinion, and predicts trends — built from scratch without external frameworks.

multi-agentagentpython +2
  • · Multi-agent Collaboration — Query/Media/Insight/Report agents collaborate through Forum mechanism
  • · Full-domain Social Monitoring — AI crawlers cover 30+ platforms including Weibo, Xiaohongshu, Douyin, Kuaishou
  • · Multimodal Content Analysis — Deep analysis of short videos, images and multimedia extracting structured info

MiroFish

71.0k · Python
Active A+

Multi-agent swarm intelligence engine that extracts seed information from the real world, constructs a high-fidelity parallel digital world with thousands of agents, and predicts future trajectories through social evolution simulations.

multi-agentswarm-intelligenceknowledge-graph +3
  • · Multi-agent swarm simulation — thousands of agents with independent personalities and long-term memory freely interact
  • · GraphRAG knowledge graph — automatically constructs a high-fidelity parallel digital world from seed materials
  • · Dynamic memory — continuously updates individual and collective memory during simulation