Breaking

Google, Microsoft, Salesforce Push A2A Agent Standard

June 20, 2026 at 09:00 EDT

AI Agents
Sales & Business
Open Source

Google, Microsoft and Salesforce are backing the open Agent2Agent (A2A) standard for making AI agents interoperate, seeking to keep enterprise apps at the center of workplace automation. The move is framed as a counter to OpenAI and Anthropic's effort to make ChatGPT and Claude the main gateway for work AI.

June 17, 2026 · Google · Microsoft · Salesforce

A "Search Engine" for the Agentic Web

Enterprise vendors jointly publish Agentic Resource Discovery (ARD) — an open spec letting AI agents dynamically discover and verify tools and other agents, keeping business apps at the center of workplace automation.

Lead backers: Google, Microsoft, Salesforce

More launch partners join the spec

Apache 2.0

Open license, draft open specification

How discovery works — before runtime connection

1 · PUBLISH

Each service exposes a catalog at /.well-known/ai-catalog.json

→

2 · INDEX

Federated registries crawl & index catalogs across the web

→

3 · CONNECT

Agents connect via native protocols — MCP, A2A, OpenAPI

The layered stack — ARD complements, not replaces

ARD — Discovery layer

Find what exists and where · Catalogs + Registries

A2A

Agent-to-agent communication

MCP

LLM access to tools & data

Trust is established via a Trust manifest with cryptographic verification — SPIFFE IDs, DIDs and X.509 certificates.

Positive reception

"The equivalent of a search engine for the agentic web." Praised as a useful discovery layer; partners highlight searching MCP servers and Skills at the scale of thousands.

Open questions

Seen by some as a move to preserve enterprise-platform dominance and check rival gateway strategies. Catalog-publishing burden and strict trust checks could add overhead; benchmarks and pricing undisclosed.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.