OpenAI just published a curated Codex use cases hub that catalogs over 60 specific, documented workflows, the company said. It’s at developers.openai.com/codex/use-cases and it’s essentially the Codex equivalent of the API Cookbook.
The hub isn’t a blog post or a single landing page. It’s a full browseable directory of Codex use cases with a consistent template for every workflow: a heading and tagline, “Best for” recommendations, a starter prompt you can copy into Codex, step-by-step instructions, and related use cases. Every page has a “Try in the Codex app” button.
The featured three
Three workflows sit at the top as the “most common Codex workflows”:
Manage your inbox. Have Codex check Gmail, pull context from Slack and Google Drive, and draft replies in your voice. It never sends anything automatically. You review first. This uses the Gmail, Slack, and Drive plugins.
Use your computer with Codex. The Computer Use plugin lets Codex click, type, and navigate apps on your Mac. It’s the fallback when no dedicated plugin exists. You start requests with @Computer and approve access.
Follow a goal. This is the /goal slash command, which gives Codex a durable objective it can work on across turns for hours unattended. You set an objective, boundaries, a validation command, and a stopping condition. Codex reports at checkpoints. It’s designed for migrations, large refactors, and prototypes.
The seven collections
The hub organizes workflows into seven themed collections:
- Productivity & Collaboration: coordinate work across apps, data, and teams
- Web development: turn designs and prompts into responsive UI
- Game development: prototype loops, UI, and gameplay faster
- Native development: build and debug iOS and macOS apps
- Production systems: navigate, refactor, and review real codebases
- Security: assess code, review changes, and remediate findings
- Life Sciences: use GPT-Rosalind for scientific research and drug discovery
Who these Codex use cases are for
The hub’s filter panel is the most revealing part. You can filter by:
- Team: Design, Engineering, Finance, Operations, Product, QA, Research, Sales
- Category: Engineering, Evaluation, Front-end, Quality, Sciences
- Workflow: Automation, Data, Integrations, Knowledge Work
- Task Type: Analysis, Code, Design, Testing, Workflow
- Platform: iOS, Mobile, macOS
Finance gets cash flow forecasting, DCF valuation modeling, and budget reviews. Sales gets meeting prep briefs and event playbooks. Research gets scRNA-seq data annotation and drug target prioritization.
OpenAI isn’t positioning Codex as a coding tool anymore. It’s positioning Codex as a universal work platform. If your job involves typing, there’s probably a documented Codex workflow for it.
Across the 60+ pages
Some of the more interesting individual use cases:
For developers: Codebase onboarding (trace request flows through unfamiliar repos), automated GitHub PR reviews, code migrations, Figma-to-code, debugging in iOS Simulator, building React Native apps with Expo.
For ops and platform teams: Dependency incident audits, vulnerability backlog remediation, deep security scans, deploying apps and websites, running verified operations workflows.
For the workflow crowd: Turn feedback into action items, turn meetings into follow-ups, turn user stories into UI mocks, coordinate new-hire onboarding, proactive teammates (a Codex thread that watches your tools and surfaces changes).
For builders: Get from idea to proof of concept (using GPT Image 2 for mockups), iterate on difficult problems (scored eval-driven optimization loop), create browser-based games, save workflows as reusable skills.
The recurring primitives
A few tools and patterns show up again and again across the workflows:
- Skills (
$skill-creator): save a workflow as a reusable skill Codex can keep on hand - Playwright (
$playwright-interactive): verify UI changes in a real browser at multiple breakpoints - ImageGen (
gpt-image-2): generate UI mockups for visual direction - Plugins: Gmail, Slack, Google Drive, Calendar, Notion, Build Web Apps, Game Studio
- <code>/goal</code>: long-running autonomous objective mode
- Automations: scheduled thread check-ins where Codex watches tools and reports signal
Tony’s Take
The most important thing about this hub isn’t the content. It’s the signal.
OpenAI could have put this in a blog post or a single markdown file. Instead, they built a structured, filterable, searchable site with 60+ individual pages, each with a starter prompt and a button to launch Codex. That’s not a documentation update. That’s an onboarding funnel.
They’re telling you: Codex isn’t for writing code. It’s for getting work done. The engineering use cases exist (and they’re well-documented), but they’re listed alongside finance modeling, sales prep, research analysis, and operations workflows. The filter by team dropdown lists Sales and Finance before it lists Engineering.
This is the same playbook OpenAI used for the API Cookbook. When they wanted developers to stop thinking of the API as “GPT wrapper” and start thinking of it as “building block for any app,” they published structured recipes. Now they’re doing the same for Codex.
The timing makes sense. Codex is a terminal-based agent in a market where Claude Code and Hermes Agent and Antigravity are all competing for the same builder audience. A use-case hub is the kind of onboarding asset that turns “I should try that tool” into “I just opened it and knew what to type.”
If you’ve been curious about Codex but never had a clear enough reason to open the terminal and try it, this hub fixes that. Pick a workflow, copy the starter prompt, and go.
