What can A2A do for you and what can you do for it
summary
A2A is an open, Protobuf-defined protocol now stewarded by the Linux Foundation that standardizes agent metadata, messages, agent card discovery and task vs message flows; it supports streaming, push notifications and MIME-typed artifacts. SDKs exist in Python and Go but lag spec v1.0.1, so expect minor mismatches. Example work includes a dozen-line Python certainty extension wired to a local aligner and a toy magic-eight-ball extension; governance runs via the Agentyc AI Foundation with public steering meetings.
AI-written summary · about
chapters (17)
- 0:00Introduction to A2A
- 1:03Early agent communication history
- 3:11LLMs and agent ontology
- 5:18A2A vs MCP or copy-paste
- 7:50A2A core concepts
- 10:18A2A for multi-agent systems
- 11:01Agent card structure
- 13:27Task and artifact communication
- 14:10Custom A2A extensions
- 16:23SDK limits and versioning
- 17:46Governance by Agentyc AI
- 19:11Open standards vs proprietary
- 20:16Minimal A2A extension
- 21:58Extension development walkthrough
- 24:04Resources and getting started
- 25:28Q&A: Running A2A locally
- 27:31Q&A: Agent capabilities
related talks
39mA2A in 45 Minutes: Building Interoperable AI Agents
16mAgent2Agent Protocol and the Future of Agent Collaboration
43mA2A in 50 Minutes: Building Interoperable AI Agents
31mA2A in 30 Minutes: Building Communicating AI Agents
27mA2A in 25 Minutes: Building AI agents that communicate
46mMCP and A2A Revolutionizing Interconnection and Collaboration of AI Agents
26mOverview of AI Agent Protocols
50mAgents That Talk Back: MCP, ACP, A2A, Oh my!