Starting Projects
Bootstrap full-stack projects in the Codex App. Set up AGENTS.md, configure local environments, and scaffold your codebase from a single thread. Start Building
Your sprint backlog has nine tickets, three of them are blockers, and your team lead just asked if anyone can “quickly” audit the auth module for security issues. You could context-switch all day and finish two of those tickets. Or you could open the Codex App, spin up parallel worktrees, and let agents work on four of them simultaneously while you focus on the architecture that actually needs a human brain.
These 20 lessons are built from real development workflows where Codex’s multi-surface approach — App, CLI, IDE extension, cloud, GitHub, Slack, and Linear — changes how you decompose and parallelize work.
Start here. These four lessons take you from an empty repository to production-ready code, using the Codex surfaces that match each task best.
Starting Projects
Bootstrap full-stack projects in the Codex App. Set up AGENTS.md, configure local environments, and scaffold your codebase from a single thread. Start Building
Codebase Analysis
Onboard to a large monorepo without reading every file. Use Codex across surfaces to map architecture, trace data flows, and understand unfamiliar code. Explore Codebases
Feature Development
Build features with parallel agents in separate worktrees. Plan locally, delegate implementation to cloud tasks, and merge results back into your branch. Build Features
Bug Hunting
Debug across the Codex App, CLI, and cloud. Reproduce locally, delegate deep investigation to cloud tasks, and tag @codex on GitHub issues to fix CI failures.
Fix Bugs
These are the bread-and-butter patterns. The workflows you reach for at 10 AM and again at 4 PM.
Testing
Generate test suites that match your existing patterns. Run tests in worktrees so the agent does not interfere with your working directory. Write Tests
Refactoring
Large-scale refactoring across hundreds of files using worktrees for isolation. Plan locally, execute in parallel, and sync results safely. Refactor Code
Code Review
AI-powered code review with /review in the CLI and App, plus @codex review on GitHub pull requests for automatic review comments.
Review Code
Documentation
Generate API docs, architecture diagrams, and README files from your codebase. Set up automations to keep docs current on a schedule. Document Code
Server-side workflows where Codex’s cloud execution and parallel agents shine.
API Development
Scaffold REST and GraphQL endpoints with validation, auth middleware, and error handling. Test endpoints in the CLI without leaving Codex’s session. Build APIs
Database Work
Design schemas, generate migrations, write complex queries, and seed test data. Delegate migration testing to cloud environments with real databases. Work with Data
Deployment
Automate deployment workflows from Dockerfiles to production. Use cloud tasks and the Codex GitHub Action to validate deployments in CI. Deploy Apps
Security Audit
Scan for vulnerabilities, audit dependencies, and generate security reports. Schedule daily security sweeps with automations. Secure Code
For the senior engineers who want to push Codex to its limits across every surface.
Performance
Profile bottlenecks, optimize database queries, and benchmark changes. Use cloud tasks with best-of-N attempts to explore optimization strategies. Optimize Performance
CI/CD Pipelines
Run Codex in GitHub Actions with the official codex-action. Automate code review, test generation, and quality gates on every pull request.
Configure Pipelines
PR Automation
Automate the entire pull request lifecycle from creation through review to merge. Tag @codex to fix CI failures and request reviews directly on GitHub.
Automate PRs
Parallel Development
Run multiple agents simultaneously in separate worktrees. Coordinate work across agents to ship an entire feature set in hours instead of days. Go Parallel
Stop doing things twice. These lessons turn manual workflows into scheduled, repeatable automation across your entire toolchain.
Codebase Modernization
Migrate legacy code at scale using cloud tasks with multiple attempts. Modernize frameworks, upgrade dependencies, and refactor patterns systematically. Modernize Code
Slack and Linear
Delegate coding tasks from Slack channels and Linear issues. Mention @Codex where your team already works and get pull requests back.
Integrate Services
Scheduled Automation
Set up recurring automations for daily code reviews, weekly dependency updates, and nightly bug hunts. Combine automations with skills for complex workflows. Automate Tasks
System Architecture
Use Codex for architectural decisions. Plan in the IDE, validate with cloud tasks, and iterate across surfaces until the design is right. Design Systems
Start with Starting Projects to get comfortable with the App, worktrees, and thread modes. Then move through Codebase Analysis, Feature Development, and Bug Hunting in order. By the end of those four, you will have a workflow that covers most daily development tasks across multiple Codex surfaces.
Jump to the lesson that matches your current bottleneck. Most experienced users find the biggest wins in Parallel Development (multi-agent worktree workflows), Scheduled Automation (set-it-and-forget-it recurring tasks), and PR Automation (GitHub integration patterns).
Focus on CI/CD Pipelines and Slack and Linear for org-wide impact. The GitHub Action and integration patterns let you embed Codex into existing team workflows without requiring everyone to install the App or CLI.