Agentic AI Best Practices

Practical habits for coding agents — OpenCode, Grok, Cursor, Claude Code, and friends

Everything on this site was built with agentic AI tools, by someone who didn't have the skills beforehand.
If you think you are not technical enough or wondered how people create such things, this is for you.
You don't need a coding background to get real value. The principles transfer across tools; setup details differ by app.

Gus Higgins  ·  Sydney, Australia  ·  LinkedIn
Agent shell
The coding-agent app that runs on your machine. Talk to it in plain language — it reads files, runs commands, calls tools, and builds real things.
+
Model access
The LLM brain behind the agent — Claude, GPT, Gemini, Grok, and more. Via an org seat, a personal subscription, or a built-in provider.
=
Real outcomes
Sites, docs, automations, and workflows — powered by your context, skills, and connectors — without writing everything yourself.
OpenCode · Grok Build · Cursor · Claude Code · Windsurf · VS Code Copilot Chat are peers — same job, different shells. This guide focuses on habits that transfer; product-specific setup still appears in some pages (we'll generalise those next). Always apply your own judgement on AI outputs, and follow any organisation AI policy that applies to you.
Start here Intro walkthrough — agentic AI from scratch A 10-section click-through story (Space / →). Mental models, demos, and watch-outs — then come back here for the full guide. →

Getting Started

Part 1 — Powering Up with Context

Part 2 — Working Habits

Part 3 — MCPs & APIs

Part 4 — Advanced / optional