Skip to content
← Inforjota
Cursos Inforjota Português

Claude Code and Codex: coding and automating with agents

From your first terminal command to a system that works on its own, safely: Claude Code, Codex CLI, Git, hooks, MCP, subagents.

Who it is for: Anyone who wants to build things with AI. Start with no coding; by the end you code with an agent at your side.

Free, no account, no ads. Your progress is saved in your browser.

Claude Code and Codex trail

Seven themed steps that point to the lessons that already exist. Each step has three stars: lessons completed, tasks done and quiz right.

  1. Using the terminal and the agent

    ★★★

  2. Asking well and testing

    ★★★

  3. Tools, context and MCP

    ★★★

  4. Subagents and models

    ★★★

  5. Skills, Git and GitHub

    ★★★

  6. Security

    ★★★

  7. Autonomy

    ★★★

Level 0: Foundations

Three short lessons for anyone who never opened a terminal: the terminal, folders and editor, reading minimal code.

  1. 1The terminal without fear · 12 min
  2. 2Folders, files and an editor (Windows, Mac and Linux) · 14 min
  3. 3Reading code, the minimum: variables, functions and errors · 13 min

Level 1: Beginner

Install, give the first task, read the diff, recover when it goes wrong, Git, secrets and permissions, context, and a site online.

  1. 1What a coding agent is (and why it is not a chat) · 12 min
  2. 2Install and sign in: Claude Code and Codex on Windows, Mac and Linux · 15 min
  3. 3Your first task: plan, diff and permissions · 15 min
  4. 4When it goes wrong: undo, restart and understand the error · 14 min
  5. 5Essential Git with the agent at your side · 15 min
  6. 6Secrets, permissions and spending caps · 15 min
  7. 7CLAUDE.md and AGENTS.md: giving the agent context · 12 min
  8. 8Project: a personal site, from zero to online · 15 min

Level 2: Intermediate

Asking well and testing, commands and skills, hooks, MCP and prompt injection, context and memory, subagents and costs. Project with API, database, tests and deploy.

  1. 1Asking an agent well: small tasks, done criteria and what a test is · 15 min
  2. 2Commands, skills and repeatable tasks · 13 min
  3. 3Hooks: automatic checks the agent cannot skip · 15 min
  4. 4MCP: connecting external tools (and prompt injection) · 15 min
  5. 5Context, compaction and memory · 14 min
  6. 6Subagents, models and costs · 15 min
  7. 7Project: an API with database, tests and deploy · 15 min

Level 3: Advanced

Agent SDK, CI/CD with agents, security in depth, multi-agent, and an autonomous system with a sandbox, spending cap and stop button.

  1. 1Agent SDK: building your own agent · 15 min
  2. 2CI/CD with agents: GitHub Actions and PR review · 15 min
  3. 3Security in depth: sandbox, supply chain and auditing · 15 min
  4. 4Multi-agent and orchestration · 15 min
  5. 5Final project: a system that works alone, with brakes · 15 min
  6. 6Appendix: cross-auditing between two model families · 12 min