Anthropic Releases ACP v2 in Draft with Enhanced Protocol Features
Key Takeaways
- ▸ACP v2 evolves beyond turn-based interactions with improved messaging and streaming tool call capabilities for more sophisticated agent behavior
- ▸More flexible permission request system provides enhanced security and fine-grained control over agent capabilities
- ▸Forward compatibility by design ensures long-term protocol stability and reduces future migration friction
Summary
Anthropic has released version 2 of its Computer Protocol (ACP) in draft form. The ACP is a foundational protocol that defines how AI agents interact with computer systems and external tools.
The v2 update introduces several significant enhancements to the protocol. Key improvements include refined messaging and streaming tool call updates that enable more sophisticated agent-system interactions, a more flexible permission request system that provides better control over agent capabilities and security, and forward compatibility by default to ensure the protocol remains stable as it evolves. The protocol redesign moves beyond the limitations of simple turn-based interactions.
To support developer adoption, Anthropic has published comprehensive v2 documentation including protocol specifications, a migration guide for existing implementations, and JSON schemas in the official repository releases (v2.0.0-alphaX). The RFD (Request for Discussion) collection provides detailed explanations of the architectural decisions behind the major changes. SDK authors can now generate code against these v2 schemas to build compliant tools.
- Open availability of JSON schemas and comprehensive migration documentation enables rapid adoption by SDK developers
