Glossary

What is the Agent2Agent (A2A) protocol?

A2A is an open standard for AI agents to discover each other and work together on tasks. How it differs from MCP, and how AP2 builds on it.

The Agent2Agent (A2A) protocol is an open standard, originated by Google and donated to the Linux Foundation, that lets AI agents built on different frameworks discover each other's capabilities and collaborate on tasks.

How does it work?

A2A treats agents as peers. Each agent publishes an Agent Card, a JSON document at a well-known path that describes its capabilities, its endpoint and the security schemes it accepts. Another agent reads that card, then sends messages to start a task, the protocol's unit of work, and receives status updates and artifacts as output. Communication runs over HTTP, with JSON-RPC and Server-Sent Events for streaming updates. A design principle is that agents stay opaque: they collaborate through declared capabilities without exposing their internal logic, memory or tools. The A2A docs describe it as complementing MCP. MCP connects one agent to tools and data; A2A connects agents to each other. AP2, the Agent Payments Protocol that Google published, is available as an extension for A2A, which is how a shopping agent and a merchant's agent can exchange signed mandates during a purchase. Google also launched an A2A x402 extension with Coinbase and others for crypto payments between agents.

In practice

You'd use A2A when your agent needs to hand work to another team's or company's agent rather than call a tool, for example a travel agent asking an airline's agent for options. Before trusting a remote agent, read its Agent Card for the authentication it requires, and treat what it returns as untrusted input, the same as web content. If payments are involved, A2A itself only carries the conversation; the payment rules come from an extension such as AP2 or from your own controls. Keep the money decision outside the conversation between agents: a request from another agent to pay should still pass the same limits, recipient checks and approvals as one from your own prompt.

What it does not mean

  • A2A isn't a payment protocol. It carries messages and tasks; AP2 and x402 add payments on top.
  • It doesn't replace MCP. An agent can use MCP for its tools and A2A to talk to other agents.
  • An Agent Card describes what an agent says it can do; it doesn't prove the agent is trustworthy.
  • Agent2Agent doesn't mean agents paying each other automatically. Any payment still needs a payment method and someone's authorisation.

Where does Veyra fit?

Veyra doesn't implement A2A or AP2. It's a remote MCP server, so any agent framework that can connect to an MCP endpoint with a bearer token can call its tools, including agents that also speak A2A. Veyra's limits apply to the payments that agent requests, whoever asked it to pay.

Veyra's spend policy for one agent endpoint: per-transaction limit, daily limit, auto-approve limit, ask-up-to limit and recipient allowlist
What this looks like in Veyra: one spend policy per agent endpoint, with a per-payment cap, a daily cap, an auto-approve band, an ask band and a recipient allowlist, all checked on the server. Try the free policy generator →

Sources

Guides that use this term

This page explains how agent payments and the related technology work. It is not financial, investment, legal or tax advice. Protocols and provider features change; check the linked official sources before relying on a detail.