Claude Code, Codex, OpenClaw and Antigravity setup
Compare four agent platforms, follow their setup walkthroughs, and learn a four-part prompt structure: goal, constraints, format and failure.
Player not loading? Watch on YouTube
This tutorial compares Claude Code, OpenAI Codex, OpenClaw and Google Antigravity, with installation and first-task walkthroughs. The speaker explains an AI agent as a model connected to tools, memory and goals through an observe, think and act loop. Examples include organizing receipts and building a simple to-do app.
The speaker favors Claude Code for complex work, Codex for existing ChatGPT subscribers, and Antigravity for visual development. These are workflow preferences, rather than measured performance results. The walkthrough describes Claude Code on Mac and Windows, including Windows ARM64, and Antigravity on Mac, Windows and Linux, with separate Apple Silicon and Intel downloads.
OpenClaw is the self-hosted, open source option discussed. Its setup covers choosing a provider, supplying an API key, selecting skills and connecting a messenger such as Telegram. Running the agent on your own computer does not establish that its model runs locally: the demonstrated configuration uses external AI providers.
The final sections explain prompt contracts: define the outcome, constraints, output format and what to do when blocked. The speaker also recommends project memory files for recurring preferences and asking the agent to record corrections. Sponsored Brevo material and a promotion for AI Master's content-production agents appear between the tutorials.
Design Blog
