ACP (Agent Communication Protocol)
Also called: Agent Communication Protocol
ACP (Agent Communication Protocol) is an earlier proposal for standardizing how AI agents communicate with each other, built on REST-style HTTP conventions.
ACP is one of several proposals that emerged alongside MCP and A2A to address agent interoperability — in ACP's case, specifically agent-to-agent communication using REST-based conventions rather than the JSON-RPC approach MCP uses. It represents an earlier or parallel attempt at some of the same problems A2A later addressed.
As of 2026, ACP has seen less adoption than MCP or A2A; it's worth knowing about as part of the landscape of agent interoperability proposals, but it isn't the protocol most current tooling and documentation defaults to. Because this space is still converging, it's reasonable to expect naming and adoption patterns to keep shifting.
For a developer evaluating protocols today, the practical takeaway is less about ACP itself and more about the pattern it represents: multiple groups have proposed different answers to "how should agents talk to each other," and REST-based versus JSON-RPC-based design is one of the axes those proposals differ on.
How it works
ACP frames agent communication in REST terms — agents exposing HTTP endpoints that other agents call to exchange messages or delegate work, following conventional REST resource/verb patterns rather than a JSON-RPC method-call model. The specifics of message format and task lifecycle differ across proposals in this space, and ACP's design choices reflect an earlier stage of the same broader problem A2A later tackled with more traction.
Common misconceptions
FAQ
What is ACP in AI agents?
Is ACP widely used?
How is ACP different from A2A?
Last checked: 2026-08-28