AI Potluck
Product / UX / Agent tools & protocols

Agent2Agent Protocol

A2A Project (Google-initiated)

Open protocol enabling communication and interoperability between opaque agentic applications. Where MCP connects agents to tools, A2A connects agents to other agents, defining task delegation, status updates, and result passing between autonomous systems. Initiated by Google with 24K GitHub stars. Still early-stage but positioned as the complementary standard to MCP for multi-agent workflows.

A2A open protocol for inter-agent interoperability; contributed by Google, hosted by the Linux Foundation (launched June 2025). Spec v1.0.1 released May 28, 2026 (v1.0 = first stable spec, adds Signed Agent Cards). SDKs in Python/Go/JS/Java/.NET/Rust. Verified live on GitHub June 2026.

Openness

5 high confidence
5.0
license
Apache-2.0(OSI)
source
public(spec + reference SDKs in 6 languages)
governance
Linux Foundation (vendor-neutral, Google-contributed)

Reference spec and SDKs are Apache-2.0 under neutral LF governance; for a protocol this is the openness ceiling.

Adoption

4 medium confidence
4.0

Protocol adoption-breadth signal (per recipe): LF press (Apr 2026 one-year mark) reports 150+ supporting organizations, in-production use at Microsoft (Azure AI Foundry, Copilot Studio), AWS (Bedrock AgentCore), Salesforce, SAP, ServiceNow, and embedding into Google Cloud. Breadth across the three major clouds + enterprise production puts effective downstream reach in the 1-10M band; this is a breadth/traction count, not a measured per-user figure.

Capability

not assessed

A2A is a wire protocol for agent interoperability; per the recipe, capability is not a meaningful axis for a protocol. The signal lives in openness and adoption-breadth.

Unchanged since 2026-06-09 (last edited, not re-checked)