Building Agents (the hard parts!)
summary
Claims that agents move work from augmentation to full automation with measured wins — 20% revenue uplift, 90% faster support and 50–75% time savings — and shows how to build them using MCP servers and Cloudflare’s agents SDK. Demonstrations use WebRTC + speech‑to‑text, LLMs for reasoning, durable objects for stateful memory, vector DBs and tool calling, plus human‑in‑the‑loop and deferred tool execution (Knock credit‑card approval example). A deploy‑button MCP server for recommendations runs, persists memory and scales without extra infra.
AI-written summary · about
chapters (12)
- 0:00Introduction and mission
- 1:52AI as a paradigm shift
- 3:40The rise of agentic workflows
- 5:24Four components of agents
- 7:52MCP standard and tool calling
- 10:24Cloudflare’s Agents SDK
- 12:08Deploying an MCP server demo
- 14:34Workflow coordination and state
- 15:36Human-in-the-loop use case
- 19:03Choosing the right model
- 19:24Client integration with MCP
- 20:50Summary and call to action
related talks
40mBuilding Agents with Agent Bricks and MCP
36mPutting AI Agents to Work: A New Era of Open Source Connectivity with MCP
01h 03mAI Coding Agents and how to code them
01h 28mAI Coding Agents and how to code them
33mAgent Experience
40mAgents in Action: Advancing Open Source AI Missions
30mAI Agents Unchained
32mHow I Build Web Applications in 2026