Build your first AI agent: what actually makes one work
Build your first AI agent: the reason-act-observe loop, choosing between sequential, reactive and planning shapes, and the ADK LoopAgent trap that burns tokens.
LLM
How Model Context Protocol (MCP) actually works
How the Model Context Protocol actually works: the three parties, what a server exposes, and the 2026 revision that made MCP stateless and dated most guides.
Connecting your database to Claude or ChatGPT without handing over the keys
Connecting a database to an LLM with one broad key is a breach waiting to happen. Learn how to build a custom MCP server with real, scoped authorization.