
Build an AI Agent From Scratch: 7 Steps, No Framework Needed
The agent loop is fifteen lines. Everything that makes an agent worth deploying is in the other 185. A working AI agent from scratch, in plain Python.

The agent loop is fifteen lines. Everything that makes an agent worth deploying is in the other 185. A working AI agent from scratch, in plain Python.

LangGraph, CrewAI, OpenAI Agents SDK, Claude Agent SDK, Microsoft and Pydantic AI compared on the criteria that actually decide it: debuggability and exit cost.

A no-code AI agent is genuinely useful and oversold about ninety percent of the time. Where the line sits, which platform to pick, and a build you can copy.

A local AI agent runs the same loop as a cloud one. The model is the only thing that changes, and that one change decides how you write everything else.

Titanic and MNIST will not get you hired - every reviewer has seen four hundred of them. 15 AI and ML projects organised by what they actually signal, plus the five rules that make any project count.

The three orchestrators disagree about something fundamental - whether to track tasks, assets or flows - and that disagreement predicts everything else. A practical comparison with a ten-minute decision guide.