AI and agents
Agent2Agent protocol (A2A)
Agent2Agent (A2A) is an open protocol, announced by Google in April 2025 and now run under the Linux Foundation, that lets AI agents built by different vendors find each other and hand work between them.
How it works in practice
MCP connects an agent to tools. A2A connects an agent to another agent. Each A2A agent publishes an agent card describing what it can do and how to reach it, and other agents send it tasks, follow progress and receive results through a shared message format. The aim is that an agent in one company’s platform can delegate to an agent in another without a custom integration.
For most companies of 1 to 1,000 people, A2A is something to watch more than something to build on today. It matters when you run several agents from different vendors, or when a partner or customer wants their agent to work with yours. The same questions apply as with any integration: who is calling, what are they allowed to ask for, and what gets logged.
A worked example
A logistics firm runs a quoting agent in one platform and a booking agent in another. With A2A, the quoting agent can pass an accepted quote to the booking agent as a task and get the booking reference back, without the team writing glue code between the two vendors.
Examples in this glossary describe typical builds for companies of 1 to 1,000 people. They are not named client results.















