Google Agent2agen’s operating protocol aims to unify the agent communications

Photo of author

By [email protected]


Join daily and weekly newsletters to obtain the latest updates and exclusive content to cover the leading artificial intelligence in the industry. Learn more


The interim operation is acquired among artificial intelligence factors slowly, as organizations begin to build networks of agents.

In the past few months, there are at least the criteria for the interventional operating agency It appeared: Antarbur Form Context Protocol (MCP) and AgntcyFrom my group led Cisco. As the importance of factors-especially those based on different frameworks and models of large language (LLMS)-Talk to each other and get a complete picture of the institution’s data gains, another new adoption protocol competes.

todayand Google A new protocol for inter -operation is called Agent2agent, or A2A, that he hopes to become a standard in industry.

Partnership with more than 50 companies, including Atlassian, Box, COHERE, Intuit, Langchain, Mongodb, Salsforce, SAP, Servicenow Betical operation language For agents and artificial intelligence applications.

In an exclusive interview, I told Rao Surapaneeni, Vice President and Director General of the Google Cloud application platform, Venturebeat that A2A is easy for agents who have different disciplines and data contract to obtain a required context.

“Everyone has a specific major because they have a logical data knot or knot, or the current user base focuses on this specified task,” said Surapanei. ))

Surapanei said part of the reason that Google has worked with more than 50 partners and customers is to build A2A to get “the ability to exchange in a ready, safe and reliable institutions.”

Based on the current standards

A2A facilitates communication between what you call Google the agent of the client and a remote agent. The customer’s agent classifies the task from the final user, and the distant agent works on the task.

In a separate blog post, Google said that A2A relies on many of the main capabilities based on the protocol.

  • Detection of abilityAgents can “announce their capabilities” through a JSON’s agent’s agent card, so that the customer agent can determine the best distant agent to complete the task.
  • Task management: Ensuring communication between agents only towards completing the requests and setting life courses for tasks.
  • cooperation: Send messages about context responses, artifacts (tasks out) or instructions.
  • User’s experience negotiatingDetermine the types of content and formats read by agents.

SURAPNENI said that Google designed A2A as an open protocol, which means that the largest open source community can contribute to the A2A project and suggest symbol updates.

“We open it as an effort that society moves and is correctly open to the source,” he said. “There is a governance council around it, but we want to be truly open and dependent on society.”

In the development of the A2A, Google focused on enabling the agents to work “in their non -structured natural methods, even when they do not share memory, tools and context.” The protocol also depends on current criteria such as HTTP and JSON, so it is easier to integrate with current technology chimneys and is virtual safe.

High inter -operating protocols

Of course, A2A is not the only inter -inter -operating protocol on the market. Agntcy, from a group of Cisco, Langchain, Galileo, Llamaindex and Glean, aims to create a standard communication method among agents. Langchain, who is also a partner in Agent2agent, phase The agent protocol. Microsoft has updated its automatic framework for help Make operating factors.

On the other hand, many companies, including Microsoft, have already adopted MCP. Even Google added MCP support through the new agent development group. SURAPANENIN confirmed that A2A will work in parallel with MCP.

“We see MCP and A2A supplementary capabilities,” said Surapaneeni. “The way we look at Agent2agent is in a higher layer of abstraction to enable applications and agents to talk to each other. So think about it as a layer stack where MCP works with LLM for tools and data.”

SURAPANI did not close the door in potential cooperation with other federations that operate the interfering of the agent’s operating protocols. He said that A2A is always open to new members, and the protocol will be a living symbol that is constantly updated based on the proposals and needs of society.

“We will look at how to comply with all protocols,” said Surapanei. “There will always be some protocol with a good idea, and we want to know how to bring all these good ideas.”

The interview is created

The organizations and artificial intelligence companies agree that the world will Operation on Multiple Amnesty International models Instead of one model governing them all, so it makes sense for agents to be built on different languages ​​and frameworks.

However, the fully coexist ecosystems requires speaking to agents from other companies. But this is easier than doing it. Industry standards often take time to keep them and require a large part of companies.

If the A2A, MCP or AGNTCY contemplates success in creating a record method for all artificial intelligence agents, regardless of those who built them or the frame they build on, there must be collective dependence and publication.

Surapanei acknowledged that even with more than 50 partners working on A2A, adoption is still at a turning point.

He said: “All these protocols will develop, especially the way the artificial intelligence rate changes, and we will find new cases of use and scenarios that must be addressed until they continue to grow.”



https://venturebeat.com/wp-content/uploads/2025/03/nuneybits_Vector_art_of_AI_agents_operating_on_the_blockchain_1e1e33ca-22c2-40f8-8f94-7baa503edc92.webp?w=1024?w=1200&strip=all
Source link

Leave a Comment