MCP at 1. The state of the protocol you love (or hate)
summary
One-year status report on MCP with concrete signals: over 36,000 MCP servers found on GitHub versus an official count of over 10,000 useful servers, major tooling (Smithery, Glama, Speak Easy, Zuplo, MCP Jam) and companies building on MCP, five clear use cases (analysis, development, automation, data APIs, actions), a Linux-hosted Agentic AI Foundation announcement, three spec changes highlighted (URL mode for credentials, client ID metadata docs, long-running tasks), plus token-cost, provenance and prompt-injection risks and Anthropic’s push for code-execution to reduce token use.
AI-written summary · about
chapters (11)
- 0:00MCP at one year
- 1:11MCP growth and adoption
- 3:20Top five MCP use cases
- 6:36Why MCP won
- 9:51MCP as universal connector
- 10:32Latest MCP spec updates
- 11:59Rapid spec changes challenges
- 13:23Security risks and prompt injection
- 14:48API provider recommendations
- 16:33MCP future and adoption
- 18:38Q&A
related talks
27mYour MCP Server is Bad (and You Should Feel Bad)
30mAPI Drift: A Hidden Crisis in MCP Servers
17mBuilding Production-Ready MCP Servers: From AI Agents to Secure Enterprise APIs.
12mBuilding Better Remote MCPs: Web3’s Answer to Auth & Monetization
31mThe MCP Gateway: A Journey from Idea to Infrastructure
39mHow to connect your systems to AI Tools with MCP Servers
01h 30mBuilding for the agentic future
20mExternal Processing, Internal Leverage: MCP Tool Calls...