Agent2Agent Protocol (A2A)

The Agent2Agent protocol (A2A) is an open standard for independent AI agents to "communicate and collaborate with each other as peers": to discover each other, negotiate how to interact and manage shared tasks. Each agent publishes an Agent Card, normally at /.well-known/agent-card.json, listing its skills, endpoints and authentication requirements.¹ Google launched it on April 9, 2025, "with support and contributions from more than 50 technology partners".²

The Linux Foundation took A2A in as a project on June 23, 2025,³ and on August 17, 2026 it became a hosted project of the Linux Foundation’s Agentic AI Foundation, alongside the Model Context Protocol.⁸ It is maintained by a technical steering committee with members from AWS, Cisco, Google, IBM Research, Microsoft, Salesforce, SAP and ServiceNow.⁴ Version 0.3.0 (July 30, 2025) added signatures to the Agent Card, and the first stable release, 1.0.0, followed on March 12, 2026, with a 1.0.1 patch on May 26, 2026.⁵ A2A 1.0 defines three protocol bindings, JSON-RPC, gRPC and HTTP+JSON/REST, and a client names its version in an A2A-Version header.¹ Agent Cards "MAY be digitally signed" with JSON Web Signature so a client can check that a card comes from the claimed provider; otherwise clients verify the server through its TLS certificate.¹ AI Agent Identity and Trust Standards sets that beside the other agent identity specifications.

At A2A's first anniversary, the Linux Foundation said more than 150 organizations supported the standard, with integrations across Google, Microsoft and AWS platforms Vendor-stated.⁶ Its JavaScript SDK was downloaded 14,889,082 times from npm in the 30 days to October 16, 2026 Measured.⁷ The specification calls A2A and the Model Context Protocol complementary: MCP connects an agent to tools and data, while A2A connects agents to each other.¹ How the Model Context Protocol Works (2026-07-28 Revision) compares the two side by side.