The multi-agent interoperability landscape has settled, but the cleanup is expensive. Agent Communication Protocol (ACP), IBM’s entry into the space, has officially merged into the Agent2Agent Protocol (A2A) under the Linux Foundation’s LF AI & Data umbrella. This consolidation ends a confusing period where three protocols—MCP, A2A, and ACP—arrived in an eighteen-month window, all promising agent interoperability. For teams that bet on ACP before the merger announcement in August 2025, the result isn’t a simple version bump; it’s a re-architecture of their coordination layer. One developer reported spending three weeks refactoring a multi-agent system because their team built on ACP, only to find it deprecated two months before they started building.

The Vertical vs. Horizontal Split

The confusion stemmed from treating these standards as interchangeable competitors. In reality, they serve distinct architectural roles. Model Context Protocol (MCP) functions as the vertical integration layer, standardizing how a single agent interacts with passive tools, databases, and APIs. By March 2026, MCP had exploded to 97 million monthly downloads, a 970x increase in eighteen months. Conversely, A2A is the horizontal coordination layer, enabling peer-to-peer task delegation between autonomous agents that possess their own reasoning and authority. A2A reached v1.0 in March 2026 and now boasts over 150 supporting organizations, including deep integrations with Google, Microsoft, and AWS. ACP, now a historical footnote, has seen its assets and expertise contributed directly to A2A, with IBM’s Kate Blair joining the A2A Technical Steering Committee.

Migration Is Not a Find-and-Replace

While a migration path exists for BeeAI users via the A2AServer adapter, it does not undo the architectural debt incurred by ACP’s assumptions. ACP utilized a brokered communication model, whereas A2A relies on peer-to-peer task delegation. Moving from one to the other requires rebuilding the coordination layer, not just swapping endpoints. The Linux Foundation framed this merger as accelerating progress toward a single, more powerful standard, but the cost was borne by early adopters who shipped on the now-deprecated protocol. The unified interoperability framework estimates that adopting the layered approach of MCP and A2A reduces integration complexity by 60–70% compared to ad-hoc solutions, but only if you build on the correct foundation from the start.

Enterprise Readiness Has Caveats

A2A’s v1.0 release introduced enterprise-grade features like Signed Agent Cards for cryptographic identity verification and multi-tenancy support. However, "A2A-compliant" is a procurement label, not a safety guarantee. The protocol solves "how to transmit," not "how to trust." Task state is self-reported by the executing agent, and while Agent Cards can carry JWS signatures, they are optional in many implementations. There is no formal registry for discovery yet, as the specification never shipped and maintainers redirected the need to adjacent specs. Salesforce’s Winter '27 Release, generally available in October 2026, signals that A2A has crossed into product packaging, but the burden of verification falls on the buyer. Organizations must validate the agents they connect, as the protocol remains agnostic to the underlying intent or reliability of the participants.

Key Takeaways

  • MCP is the standard for tool access (vertical), with 41% of software organizations reporting it in production as of July 2026.
  • A2A is the standard for agent-to-agent coordination (horizontal), supported by 150+ organizations and embedded in Azure AI Foundry and Amazon Bedrock.
  • ACP is deprecated and merged into A2A; teams built on it face re-architecture, not just migration.
  • "A2A-compliant" certifies connection language, not safety or identity verification, which must be built by the adopter.
  • The payments layer remains fragmented, with practitioners advising to treat it as a piloted capability behind an abstraction, not a foundational commitment.

The Bottom Line

Stop treating "compliant" as a synonym for "safe." The protocol war is over, but the real work—building verification layers and managing context window scaling issues in MCP—is just beginning for teams shipping in production.