Model Context Protocol (MCP) and agentic skills
summary
A practical teardown of Model Context Protocol shows how bad skill descriptions and sloppy names stop tool calls cold, with parameter schemas and activation logic being the usual failure modes; a one-line MCP tool and Java TCP/Spring MVC endpoints get wired to Claude-style models and lift activation from 30%→90% activation (a ~3x uplift) after renaming and better JSON schema; notes compare RPC-over-HTTP to plain HTTP+JSON and point to MCP naming and eval artifacts.
AI-written summary · about
chapters (9)
- 0:00Introduction and welcome
- 1:07Why context matters for AI
- 2:09What is Model Context Protocol
- 3:36MCP servers and tools
- 5:49Testing MCP tool activation
- 7:183x uplift with better skills
- 8:48Evaluating MCP tools across models
- 9:51Writing precise MCP skills
- 10:33Call to action for MCP skills
related talks
33mThe Model Context Protocol: Connecting AI to Everything
26mBuilding AI-Enabled Spring Applications with Model Context Protocol (MCP)
45mModel Context Protocol (MCP): The future of user interactions?
50mBeyond local tools: Dive into the Model Context Protocol (MCP)
32mMCP for dummies: Feed more context to LLM
44mBeyond the Prompt: How MCP Connects Language Models to Real Tools
32m30 minutes to understand MCP (Model Context Protocol)
46mMCP for LLM Agents, Infrastructure APIs & Data Services