Hold
IBM/BeeAI's REST-based agent communication protocol — functional but merging into A2A under the Linux Foundation. New projects should target A2A directly.
Why It Matters
ACP was a promising, clean REST-based approach to agent communication. However, the team has announced it is merging into A2A under the Linux Foundation umbrella. If you have existing ACP integrations, they still work. But the writing is on the wall — the community is consolidating around A2A, and investing further in ACP-specific work is wasted effort.
Strengths
- Clean REST-based design that was easy to implement and reason about
- Existing integrations continue to function — no immediate migration pressure
- Concepts and patterns carry over well to A2A, so learning wasn't wasted
- The merge means ACP's best ideas will influence A2A's evolution
Limitations
- Officially merging into A2A — no independent future
- New features and fixes will land in A2A, not ACP
- Shrinking community as developers migrate to A2A
- Starting new projects on ACP creates guaranteed migration debt
Risks
- ACP is a dead protocol walking — continuing to invest in it is throwing good money after bad
- The "merge into A2A" narrative is diplomatic framing; in practice, A2A won and ACP lost
- IBM's track record with open-source projects (see: OpenWhisk, etc.) suggests they'll quietly deprioritize once the PR value fades
- Migration from ACP to A2A is described as "concepts carry over" but actual code migration is non-trivial