Building Interoperable Agents: When A2A meets MCP
summary
A working demo named Duct Tape platform generates repair tutorials from an image by chaining an orchestrator, gatekeeper, researcher, judge and content‑creator agents; the researcher calls an MCP tool to deterministically find a YouTube video and the content creator assembles step‑by‑step pages. Architecture uses an ADK wrapper over LangChain, Gemini 1.5 (80K) as the model, MCP servers for typed tool schemas and the agent‑to‑agent protocol for cross‑silo calls. Code and deployments (Cloud Run) are open source Apache 2.0 and the repo already exceeds 20,000 stars.
AI-written summary · about
chapters (13)
- 0:00Interoperable agents intro
- 1:14Duct Tape platform demo
- 5:58What are agents
- 8:28Gatekeeper agent role
- 10:55Agent Development Kit
- 13:45Open source in agents
- 14:30Model Context Protocol
- 17:25MCP tools integration
- 19:54Agent-to-Agent protocol
- 23:28System architecture
- 25:17Multi-agent principles
- 26:41Deploying agents reliably
- 27:22Code and resources access
related talks
36mPutting AI Agents to Work: A New Era of Open Source Connectivity with MCP
40mBuilding Agents with Agent Bricks and MCP
46mMCP and A2A Revolutionizing Interconnection and Collaboration of AI Agents
01h 03mAI Coding Agents and how to code them
01h 28mAI Coding Agents and how to code them
50mAgents That Talk Back: MCP, ACP, A2A, Oh my!
02h 59mStandards and AI Agents: an Overview of MCP, A2A, ADK and More
40mAgents in Action: Advancing Open Source AI Missions