MCP deep dive (+ claude code?)

​MCP (Model Context Protocol) is Anthropic’s proposal for a standardized protocol for AI agents to call external tools. It’s supported by the Claude desktop app, Claude Code, Cursor, and others, and seems to be gaining broad adoption.

​Technically, it’s basically tool use with some extra plumbing to make it easier to plug arbitrary tools together. Right now, it’s local-only, so, e.g., Claude Desktop can connect to a server running locally that provides arbitrary services. Remote tool use is coming in the future.

​This event is an informal schelling point to discuss MCP, experiences, tools that use it, AI agent tool use in general, Anthropic’s strategic direction with this, etc.