A2A (Agent-to-Agent Protocol)
An open protocol developed by Google that enables AI agents built by different vendors or in different frameworks to securely work together, exchange information, and coordinate actions without requiring proprietary integration
As defined in Google Agentspace JUST WON the AI Game...
Where it comes up · 1
Most important claims · 10 of 13Search all claims for “A2A (Agent-to-Agent Protocol)” →
- Google has announced Agent Space, a platform that creates an ecosystem where businesses and developers can create agents that seamlessly work together, coupled with an open A2A (agent-to-agent) protocol for developing interoperable agents.Google Agentspace JUST WON the AI Game...
- The A2A protocol enables agents to coordinate complex workflows involving multiple specialized agents, each with narrow, well-defined capabilities, sequencing their work and managing dependencies (e.g., candidate sourcing → interviews → background checks with different agents for US vs international candidates).Google Agentspace JUST WON the AI Game...
- An agent in the A2A ecosystem can select between multiple available agents providing the same service based on past interaction experience and agent reputation, choosing the agent with the 'best experience in past interactions'.Google Agentspace JUST WON the AI Game...
- The A2A protocol in demonstrated hiring scenario supports multi-week task execution with human involvement, where an agent coordinates candidate sourcing, conducts interviews over multiple days, initiates background checks, tracks completion status, and provides user updates on task progress.Google Agentspace JUST WON the AI Game...
- The A2A protocol design follows principles including: embracing agent capabilities by enabling agents to collaborate in unstructured modalities even without shared memory, tools, or context; building on existing standards for easy integration; being secure by default with enterprise-grade authentication; supporting long-running tasks from quick operations to days-long processes; and being modality-agnostic (text, audio, video, streaming).Google Agentspace JUST WON the AI Game...
- Agents in the A2A registry advertise their capabilities using simple JSON format, enabling capability discovery where agents can list specific capabilities (e.g., 'US-based background checks') and clients can select agents based on matching these declared capabilities.Google Agentspace JUST WON the AI Game...
- The A2A protocol architecture enables task-oriented communication where the output of a completed task is called an 'artifact', and this architecture is foundational to the Agent Development Kit (ADK) which allows developers to build production-ready multi-agent applications with greater flexibility and control.Google Agentspace JUST WON the AI Game...
- Anthropic's Model Context Protocol (MCP) allows large language models to communicate with other software; the A2A protocol complements MCP by enabling agent-to-agent communication at a higher level of abstraction than model-to-system communication.Google Agentspace JUST WON the AI Game...
- In the A2A protocol architecture, the client agent represents the user and is responsible for secure collaboration, task and state management, user experience negotiation, and capability discovery with remote agents, while the remote agent performs the delegated work and communicates results back.Google Agentspace JUST WON the AI Game...
- As the A2A agent ecosystem grows with more capable agents added to it, it will become easy to monetize the platform through a single chat interface that provides access to millions of agents, with any agent capable of performing a task being plugged in and made available to users.Google Agentspace JUST WON the AI Game...