Google offers open protocol for AI agent connectivity
The idea of having AI agents coordinate activities to achieve a task requires a standard protocol for sharing queries, answers and admin tasks
With support from technology partners, Google has introduced an open agent-to-agent protocol to enable artificial intelligence (AI) agents to communicate securely with one another.
Google said it has drawn on internal expertise in scaling agentic systems to design the A2A protocol to address the challenges it identified in deploying large-scale, multi-agent systems for customers.
Google said A2A enables developers to build agents capable of connecting with any other agent built using the same protocol. “Businesses benefit from a standardised method for managing their agents across diverse platforms and cloud environments,” the company said on its developer blog.
The protocol provides a mechanism for agents to advertise their capabilities using an Agent Card, which is submitted in the machine-readable JavaScript object notation format. Google said this enables a client agent to identify the best agent to perform a task. Agents can send each other messages to communicate context, replies, artefacts or user instructions.
As an example, Google said the process of hiring a software engineer can be significantly simplified using agents that support A2A collaboration.
A hiring manager can task an agent to find candidates matching a job listing, location and skillset. The agent then interacts with other specialised agents to source potential candidates. The user receives these suggestions and can then direct the agent to schedule further interviews, which, according to Google, helps to streamline the candidate sourcing process.
SAP is one of the companies that announced support for A2A. It said that in the case of a customer dispute coming in over Google’s Gmail, rather than toggling between tools, a contact centre agent can invoke SAP’s AI copilot, Joule, directly from the email.
Read more about AI agents
- DeepMind founder warns of compounding AI agent errors: Artificial general intelligence may be years away, but in the meantime, organisations are being urged to adopt agent technology.
- A technical guide to agentic AI workflows: Implemented correctly, agentic AI workflows can make business process management more responsive, flexible and autonomous.
In SAP’s example, Joule, acting as an agent orchestrator, initiates a dispute resolution process, engaging another Google agent that connects to Google BigQuery, where relevant transactional warehouse data resides. SAP said the agents validate the issue, retrieve insights and recommend a resolution, all without manual system switching, data reconciliation or context loss.
Service providers supporting A2A include Accenture, BCG, Capgemini, Cognizant, Deloitte, HCLTech, Infosys, KPMG, McKinsey, PwC, TCS and Wipro.
“As agentic AI evolves, seamless handling of multi-modal data – text, voice, enterprise videos and images – becomes paramount,” said Miku Jha, director of AI, machine learning and generative AI at Google Cloud.
“This introduces significant challenges for agent interoperability. An open protocol like A2A is therefore indispensable, providing the necessary framework and flexibility for agents to effectively communicate and collaborate across these diverse modalities. Multi-modality is not simply a capability; it is a foundational requirement driving the next generation of interconnected agentic systems.”
Other software providers supporting A2A include Atlassian, Box, MongoDB, Neo4j, New Relic, Salesforce and ServiceNow. While Microsoft copilot agents do not appear to be directly supporting the A2A protocol, both Microsoft and Google work with Anthropic’s Model Context Protocol (MCP), which provides tools and context to agents.
Last month, Microsoft announced integration with MCP, which provides a mechanism for users of Copilot Studio to connect to existing knowledge servers and application programming interfaces.
Originally published at ECT News
