Assessing
Started digging into MCP — Anthropic's protocol for connecting external tools to AI agents. The idea is simple and powerful: you write a server, describe your tools via a schema, and the agent figures out on its own when and how to call them.
I'm trying to write my own MCP servers so that agents can interact with the tools I've built and with internal systems — without having to explain to the agent by hand, every single time, what lives where.
Still scouting: curious to see how this takes root in real working scenarios and how stable the protocol stays as the ecosystem grows.