Composable AI: Building Next-Gen AI Agents with MCP
summary
Explains how the Model Context Protocol (MCP) turns APIs into reusable tool servers so agents call tools (via function calling) instead of custom integrations, using MCP servers for Postgres, Snowflake, Slack, browser and VS Code, standard transports like stdio and SSE (HTTP streaming coming), and authentication by forwarding bearer tokens. Claims thousands of MCP servers exist, that 50–100 tools is the practical sweet spot, and gives a real example: a social manager product built from 29 MCP servers that runs in production and charges users.
AI-written summary · about
chapters (8)
- 0:00Composable AI and MCP
- 2:38AI agent definition and tools
- 7:15Need for MCP and how it works
- 11:54Building MCP servers and AI
- 15:06Reusing APIs with MCP
- 17:59Scaling agents with tools
- 21:10Agent protocol and future standards
- 23:22Q&A on MCP
related talks
33mThe Model Context Protocol: Connecting AI to Everything
37mLetting AI Interface with Your App with MCPs
36mPutting AI Agents to Work: A New Era of Open Source Connectivity with MCP
46mMCP for LLM Agents, Infrastructure APIs & Data Services
29mEmpowering Your Product with Agentic AI
50mBeyond local tools: Dive into the Model Context Protocol (MCP)
35mMCP in Action: Real-World Use Cases & Design Patterns
42mMCP in Practice: One App to Understand It All