A2M, MCP, A2A and AP2

A2M is meant to be used alongside existing agent protocols, not instead of them.

Protocol	Connects an agent to	A2M’s relationship
MCP	Tools, data and prompts	A2M’s first binding is a set of MCP tools behind MCP’s OAuth authorization. A2M adds what MCP leaves to each server: mandates, signed decisions, execution state and evidence for orders.
A2A	Other agents	Complementary. An A2A agent that wants to trade would still need an A2M authorizer and executor. No A2A binding is defined in v0.1.
AP2	Payments on a user’s behalf	Closest in spirit: both use a user-granted mandate to bound what an agent may do. AP2 is about payments; A2M is about orders in financial markets. A2M v0.1 does not use AP2 objects and claims no compatibility.

In short: MCP is how an agent uses tools, A2A is how agents talk to each other, A2M is how an agent’s trade gets authorized and executed.

Name collision

The acronym A2M is also used by an unrelated open-source project, the Agent-to-Memory protocol, which defines shared memory for agents. This site is about Agent-to-Market. Where the context could be ambiguous, use the full name or agent2market.