A2A protocol
The A2A (Agent2Agent) protocol is an open standard that lets AI agents built by different companies talk to each other: find out what another agent can do, give it a task and get the results back. Google started it, and it is now an open project under the Linux Foundation.
How it works
Each agent publishes an “Agent Card”, a short machine-readable description of what it is, what it can do and how to reach it. A client agent reads the card and sends a task. The remote agent works on it and reports progress and results, which might take seconds or days. The remote agent doesn’t have to reveal its internal tools, memory or prompts. It only exposes the work it agrees to do.
A2A vs. MCP
The two often come up together, and they solve different problems. MCP connects an agent to tools and data, like a database, a CRM or a file store. A2A connects an agent to another agent, which has its own judgment and may belong to a different company. One agent might use MCP to reach its tools and A2A to delegate part of a job to someone else’s agent.
Why it matters
As companies deploy more agents, more jobs cross boundaries. A procurement agent needs something from a supplier’s sales agent, or an internal agent has to hand work to one run by another team. Without a shared protocol, each pair of agents needs its own custom integration.
A protocol only covers the plumbing, though. It doesn’t settle which agents should trust each other, who each agent is acting for, or when a person needs to approve something. Those questions still need answers in whatever system the agents run in.
How elvex relates
elvex takes a different route to agents working together. Each person’s agent connects to elvex over MCP and joins shared threads, acting only for its own person, and agents read each other’s messages as information, never as instructions.