a2a-go

Overview

A2A-Go provides a comprehensive framework for building scalable, distributed agentic AI systems using Go. It implements the Agent-to-Agent protocol from Google and integrates with Model Context Protocol (MCP) for standardized tool interaction. The framework features a microservice architecture where agents and tools operate as independent services with built-in VPN capabilities. It includes multiple specialized agents (Manager, Planner, Researcher, Developer) and a catalog service for service discovery.

Try asking AI

After installing, here are 5 things you can ask your AI assistant:

• send_task

  Send a new task to an agent

  Try:Create a new task for the research agentTry:Send a task to the developer agent to write code
• get_task

  Retrieve a task by ID

  Try:Get the status of task t123Try:Retrieve task details for task t456
• cancel_task

  Cancel a task by ID

  Try:Cancel task t123Try:Stop the running task t456
• stream_task

  Stream the results of a task

  Try:Stream results for task t123Try:Show live updates for task t456
• list_prompts

  Retrieve a list of prompts from an agent

  Try:List all available promptsTry:Show me the prompts stored in the agent
• get_prompt

  Retrieve a prompt by ID

  Try:Get the greeting promptTry:Retrieve prompt with name 'summary'
• set_prompt

  Create or update a prompt

  Try:Create a new prompt called 'analysis'Try:Update the existing prompt 'summary'
• delete_prompt

  Delete a prompt by ID

  Try:Delete the greeting promptTry:Remove prompt with name 'summary'
• tool_calling

  Call tools and receive the results

  Try:Call the browser tool to search the webTry:Use the GitHub tool to create an issue
• sampling

  Sample a task from an agent

  Try:Sample a task from the research agentTry:Get a sample task from the developer agent
• roots

  Get the root task for a task

  Try:Get the root task for t123Try:Show the origin task for t456
• browser

  Headless browser automation with Rod

  Try:Open a website and capture a screenshotTry:Navigate to google.com and search for 'AI'

Comparable tools

Installation

git clone https://github.com/theapemachine/a2a-go.git
cd a2a-go
go build -o a2a-go main.go
./a2a-go --help

docker build -t theapemachine/a2a-go:latest .
docker run -it --rm theapemachine/a2a-go:latest ui

make server  # Start all services
make client  # Run the Terminal UI client

{
  "mcpServers": {
    "a2a-go": {
      "command": "docker",
      "args": ["run", "-it", "--rm", "theapemachine/a2a-go:latest", "ui"]
    }
  }
}

FAQ

What is the difference between A2A protocol and MCP?

A2A (Agent-to-Agent) is Google's protocol for communication between AI agents, while MCP (Model Context Protocol) is Anthropic's protocol for standardized tool interaction and data exchange. A2A-Go implements both protocols and enables interoperability between them.

How does the VPN functionality work?

A2A-Go includes a built-in VPN solution using WireGuard specifically for the browser tool. It provides SOCKS5 proxy support and network context switching, allowing the browser to operate from different network endpoints with custom network stacks using gVisor.

Compare a2a-go with