Processing…
AI & Architecture
After MCP to connect AIs to tools and data, A2A standardizes exchanges between agents from different technologies and organizations.
Agent2Agent (A2A) protocol joined the Agentic AI Foundation as a hosted project on August 17, 2026. Already under Linux Foundation governance since 2025, A2A has a common home with other bricks in the open agent ecosystem.
A2A defines a common way for agents from different suppliers or frameworks to discover each other, delegate tasks and exchange results. Version 1.0, published in March 2026, notably strengthened the versioning, multi-lease and identity of agents.
MCP mainly standardizes the connection of an AI to its tools and data. Rather, A2A concerns communication between independent agents. Both protocols can therefore coexist in an architecture where each component retains a clear role.
For a company, this is gradually paving the way for specialised systems that are less dependent on a single supplier. It does not justify transforming each application into a multi-agent system: business gain, security, traceability and costs remain the first criteria.
Read the official announcement of the AAIF · Discover the applied AI approach →
AGENTIC ARCHITECTURE
Protocols, responsibilities, supervision and error recovery must form a coherent system. We can design that foundation with you.
Enter at least 2 characters to start searching.
