# Codex (CLI + cloud) > Codex CLI replaces the recommended stack's Claude Code. Then the differentiator: Codex Cloud takes a ticket from the queue, opens a PR autonomously, and pings you when it's ready for review. Pair it with Claude reading the diff and the human reviewer mostly approves clean diffs. Source: https://aistack.sh/stack/ship-code-codex Category: Code · Workflow Level: Intermediate Last verified by a human: 2026-05-05 ## The tools ### Codex — CLI for active work + Cloud for backlog PRs - Pricing: Free · $20/mo ChatGPT Plus · $100/mo Pro 5x · $200/mo Pro 20x · $25/user/mo Business · API metered - Site: https://openai.com/codex - Why this one: Codex CLI handles the same role Claude Code does: terminal agent with file system access. Codex Cloud is the ChatGPT-bundled cloud lane that takes a task and opens a PR while you do something else. Use both: hands-on for the work you're thinking about, cloud for the cleanup tickets in the queue. - Swap in instead: Claude Code, Cursor - More: https://aistack.sh/tool/codex ### Claude — Cross-vendor diff review - Pricing: Free · $20/mo Pro · $100/mo Max 5x · $200/mo Max 20x · Sonnet API $2/$10 per M tokens (input/output) - Site: https://claude.ai - Why this one: Cloud PRs especially want a second model reading the diff. Claude is a different lineage than Codex, so it catches things Codex's self-review wouldn't (different priors on idiomatic code, different blind spots). Paste the PR diff + the ticket, ask for a structured review. - Swap in instead: CodeRabbit - More: https://aistack.sh/tool/claude ## Monthly cost ### $0/mo — Solo OSS maintainer (small) - Codex: Free tier - Claude: Free tier ### $40/mo — Indie dev, daily PRs, ChatGPT Plus + Claude Pro (medium) - Codex: $20 (bundled with ChatGPT Plus) - Claude: $20 Pro ### $320/mo — Cloud-leaning team, ChatGPT Pro plans (heavy) - Codex: $200 ChatGPT Pro - Claude: $100 (5× $20 Pro) - + misc: $20 (extra cloud-task budget) ## Workflow ### 1. Decide: hands-on or backlog? If you're already thinking about the change, run the CLI. If it's a clear ticket from the backlog (refactor, dep bump, doc update), send it to Cloud and move on. ### 2. Hands-on with Codex CLI (Codex) `codex` in your repo. Same shape as Claude Code: scope, plan, implement, run tests. **Prompt: Plan-and-implement from a ticket (Codex CLI)** ``` Plan first, then implement. Ask 1 to 2 clarifying questions before you start writing if anything is genuinely ambiguous. Ticket / problem: """ {{paste ticket}} """ Constraints: - Smallest viable diff. - Match the existing code style. Read nearby files before introducing a new pattern. - Run the project's test command. Don't report done if it fails. - Treat AGENTS.md / CLAUDE.md / .cursorrules at the repo root as authoritative. When done: print the file list, the test result, and a 2-line summary I can paste as the commit message. ``` ### 3. Backlog tickets via Codex Cloud (Codex) From ChatGPT, hand off the ticket with the repo selected. Cloud reads the repo, plans, opens a draft PR, and pings you. Read the PR description first; reject and retry if scope drifted. ### 4. Cross-vendor review with Claude (Claude) Paste the PR diff + the original ticket into Claude. Ask for a structured review (coverage, edge cases, regressions, naming, idiomatic-code drift). For Cloud PRs especially, this catches drift the autonomous lane introduced. A different model lineage notices things the same vendor's self-review wouldn't. **Prompt: Cross-vendor PR review** ``` Review the PR below as a skeptical senior engineer who didn't write it. Original ticket: """ {{paste ticket}} """ Diff: """ {{paste git diff}} """ Output: 1. Does the diff actually solve the ticket? Anything missing? 2. Edge cases or failure modes the implementation doesn't handle (pick the 3 most likely). 3. Naming / idiomatic-code drift. Flag anything that looks like cross-vendor pattern leakage. 4. Specific lines I should re-read. 5. Verdict: ship / fix list / scrap and re-prompt. Be specific (line refs, not vibes). Don't pad. ``` ### 5. Bounce findings back to Codex (Codex) For CLI work: paste Claude's findings into the same Codex session. For Cloud PRs: kick the same task back via the Cloud `Address review feedback` action with the findings as the new prompt. ## What it produced **Two-engineer startup, mixed-mode week** Half the PRs (the ones requiring product judgment) went through Codex CLI. The other half (dep updates, type-error cleanups, doc touch-ups) ran on Codex Cloud while the engineers worked on real product. The Cloud lane shipped 22 PRs in one week with 3 needing a retry for scope drift. ## Pitfalls - **Codex Cloud scope drift** — Autonomous lanes occasionally rewrite the world. Always read the PR description before the diff; if the description says 'also refactored…', kill the PR and retry with a tighter ticket. - **Cloud-task budget burn** — Cloud tasks are metered separately from the chat subscription. A team that sends 30 cloud tasks/day will exceed the bundled budget within a week. Track it. - **Conflating CLI tells with Cloud tells** — Codex CLI is supervised; Cloud isn't. The same prompt works very differently in the two lanes. Don't copy a CLI prompt into Cloud and expect the same result. ## Verifiable artifacts - [OpenAI Codex CLI — live demo (6:20)](https://www.youtube.com/watch?v=FUq9qRwrDrI) — Recorded by OpenAI on the OpenAI YouTube channel. Vendor-authored, but a real terminal session against a real codebase — useful for seeing the actual prompt-to-diff loop before committing to the workflow. ## Other ways to do this - [Ship code with AI](https://aistack.sh/stack/ship-code-with-ai): Plan in Claude chat, write in Claude Code, review with CodeRabbit — $50/mo - [Antigravity (Google)](https://aistack.sh/stack/ship-code-antigravity): Google's AI IDE: Gemini 2M-context inside the editor — $40/mo - [Claude Code only](https://aistack.sh/stack/ship-code-claude-code): Pure-CLI: scope, plan, write, fix all inside Claude Code — $50/mo - [Cursor (IDE-first)](https://aistack.sh/stack/ship-code-cursor): AI inside the editor, not in a chat tab or terminal — $20/mo - [Windsurf (Devin Local)](https://aistack.sh/stack/ship-code-windsurf): Cognition's IDE with the Devin Local autonomous-edit agent — $15/mo --- Curated by alex-w on aistack.sh. Last updated 2026-05-05.