I build Morphy, so read this with that in mind. Since July, OpenAI's Codex app lives inside the ChatGPT desktop app, and if OpenAI is the only AI you use, it's very good. Here's where it's the better pick, and where Morphy is.
What are they?
The Codex app is now a view in the ChatGPT desktop app, next to Chat and Work. OpenAI merged them on July 9, 2026, on macOS and Windows (from the Microsoft Store), with a Linux preview since August. Each chat works either in your project folder or in its own git worktree, and Hand off moves a chat and its code between the two. On Plus and up, tasks can also run in the cloud. Around that: a review pane, a terminal per chat, a built-in browser, skills, MCP servers, plugins and scheduled tasks. It runs OpenAI's models, with GPT-6.1 Sol as the recommended one.
Morphy is a desktop app with a canvas: every chat is an agent, and you can put as many of them as you like next to each other. A Codex chat in Morphy runs the Codex command-line tool with your ChatGPT sign-in, next to chats on Claude Code, Copilot, Antigravity, Grok, OpenCode, OpenRouter, Ollama or (on Pro) any OpenAI-compatible API. Each project is a group on the canvas with its own folder, focus mode tiles the agents you're working with, and Morphy's own assistant can start and check on the others for you.
How do they compare?
| Codex (ChatGPT app) | Morphy | |
|---|---|---|
| Runs on | macOS, Windows, Linux (preview) | Windows 10/11, macOS 11+, Linux (beta) |
| Agents | Built around OpenAI's models; other providers by editing its config (they must offer OpenAI's Responses API); local models are documented for the CLI | Claude Code, Codex, Copilot, Antigravity (Gemini), Grok, OpenCode, OpenRouter, Ollama, and (on Pro) any OpenAI-compatible API |
| Keeps agents apart | Git worktrees for independent chats | None built in: chats in a project share its folder (point a chat at another folder yourself) |
| Price | Included in ChatGPT plans: Free and Go ($8) get a smaller model in the app only; Plus $20 adds the cloud, CLI and code review; Pro from $100 a month | Free plan (3 agents at a time, 2 projects); Pro $5/mo (early price); you use your own AI plans |
| Open source | The app no; the Codex CLI and engine yes (Apache-2.0) | No |
| Best for | Codex users who want parallel chats from OpenAI itself | Seeing many agents, on any AI, at once |
And feature by feature:
| Codex app | Morphy | |
|---|---|---|
| Agents visible at once | Chats in a sidebar of projects; pop a chat out into its own window, and an Activity view lists the running, waiting and unread ones | As many as fit the canvas; focus views tile any number |
| Agents working at once | No set number; your ChatGPT plan's usage is the limit | 3 on the free plan, any number on Pro |
| Other AIs | OpenAI's models; other providers by editing config.toml (they must offer the Responses API), plus Amazon Bedrock built in | Claude Code, Copilot, Antigravity, Grok, OpenCode, OpenRouter, Ollama, picked from a menu |
| Codex on free or local models | Local models are documented for the command line (--oss) | OpenRouter's free models and Ollama run on Codex's own agent, set up in Settings |
| Git worktree per chat | Yes, with Hand off between your folder and a worktree | No; agents in a project share its folder |
| Diff review and line comments | Yes, plus a pull request inbox, reviews and Fix for failing CI checks | No; review in your editor or with git |
| Cloud tasks | Yes, on Plus and up, in OpenAI's virtual machines | No; agents run on your computer while Morphy is open |
| Built-in terminal | One per chat | As many as you like per project; on Pro, a chat can switch to Codex's own terminal screen |
| Built-in browser | Yes, with comments on your local pages; computer use on macOS and Windows | Yes, with phone and tablet sizes; on Pro, click an element on your local site to send it to a chat |
| Phone | The ChatGPT app on iOS and Android, from anywhere (the computer must be a Mac or a Windows PC) | On Pro, Morphy Remote in your phone's browser, over your own Wi-Fi |
| Scheduled tasks | Yes, while the app is running | One message, sent after a delay; Automations (beta, Pro) for multi-step jobs you start yourself |
| Price | Included in ChatGPT: Free and Go get GPT-6 Luna in the app only; Plus ($20) adds the cloud, the command line and code review | Free plan, or Pro at $5 a month (an early price); you pay for the AIs you use |
Where is the Codex app better?
It's the more complete tool for an OpenAI-only workflow, and it's built by the people who make Codex.
- Review is built in. The review pane shows the diff for your unstaged changes, a commit, a branch or the last turn, and you can comment on lines, stage or revert a hunk, and run
/review. A pull request inbox lets you submit a review, and Fix sends failing CI checks back to the chat. Morphy has none of that: you review in your own editor or with git. - Agents can't overwrite each other. Put a chat in a worktree and it gets its own copy of the project on its own branch, and Hand off brings it back to your folder. In Morphy, agents in one project share its folder, so give each its own folder if they might touch the same files.
- It keeps going without you. On Plus and up, cloud tasks run in OpenAI's virtual machines and keep going while your computer sleeps, and the ChatGPT phone app reaches your Mac or PC from anywhere. Morphy's agents run only while Morphy is open, and Morphy Remote (on Pro) works on your own Wi-Fi.
- More automation. Scheduled tasks, computer use on macOS and Windows, plugins, and memories that stay on your computer.
- Nothing extra to pay. It comes with your ChatGPT plan, and Free and Go can try it with GPT-6 Luna.
Where is Morphy better?
- Claude Code next to Codex. The Codex app runs OpenAI's models. In Morphy, a Codex chat and a Claude Code chat can work on the same project side by side, with Copilot, Gemini or Grok next to them, and you can compare what they come back with. When your ChatGPT plan's five-hour window runs out, the next chat runs on Claude or a free model.
- You see the whole crowd. In the Codex app, chats sit in a sidebar, you can pop one out into its own window, and an Activity view lists the ones running, waiting or unread. A canvas shows every agent at once, across projects and AIs, each with its status (working, or waiting for you), so you notice the one that needs an answer. On the free plan, 3 of them work at a time; Pro takes that limit off.
- Codex on free and local models, without a config file. In Morphy, OpenRouter's free models and Ollama on your own computer run on Codex's own agent, connected in Settings. On Pro, a custom provider can be any OpenAI-compatible API: Morphy translates for services that only offer the older chat completions, like DeepSeek's or Mistral's own APIs. In the Codex app, another provider means editing
config.toml, it must offer the Responses API, and local models are documented for the command line only. - An assistant that runs the others. Ask Morphy to "start three Codex chats in my website folder and have them review the checkout" and it does, on any AI and in any folder; ask it later which are done and it checks every chat.
- Things that stay on the canvas. Ask for a dashboard, a tracker or a tool and it appears as a live card next to your agents.
Do they share the same limits?
Yes. A Codex chat in Morphy signs in with your ChatGPT account, so it uses the same plan as the Codex app. On Plus, usage is counted in five-hour windows, and weekly limits may apply too; Pro has no five-hour limit. Five Codex agents use your plan five times as fast in either app. Morphy shows how much of your Codex plan you've used and when it resets (in Settings → Plan usage), and lets the next chat run somewhere else when it's used up.
What about privacy?
In both apps, Codex's work goes to OpenAI, so OpenAI's rules apply either way. On Free, Plus and Pro, OpenAI may train on it unless you turn off Improve the model for everyone in ChatGPT's data controls; Codex's Include environments setting is separate. Business, Enterprise and Edu aren't used for training by default. The Codex app's cloud tasks run with a copy of your code on OpenAI's machines.
Morphy adds a Google sign-in, a once-a-day "opened" note and a check of which plan you're on, and has no third-party analytics. Your messages go straight from your computer to the AI you picked.
Can you use both?
Yes. Both use your ChatGPT sign-in, and Morphy can point a chat at any folder. A setup that works well:
- Run the agents in Morphy, on whichever AI suits each task: Codex for one job, Claude Code for another, a free model for the small ones.
- Open the project in the Codex app when it's time to review the diff, comment on lines and open the pull request, or to send a long task to the cloud.
Which should you pick?
- Only OpenAI, and you care most about review and pull requests: the Codex app.
- Codex and Claude Code together: Morphy. See Codex vs Claude Code for how the two agents compare.
- You want agents to keep working with your computer asleep: the Codex app, with cloud tasks on Plus or higher.
- No paid AI plan yet: the Codex app on Free or Go, or Morphy with OpenCode's or OpenRouter's free models, or Ollama. See Connect your AI.
- Only Claude: see Claude Code desktop vs Morphy, or every option in our roundup.
References
- OpenAI: What's new in the ChatGPT app (the Codex merge) and the changelog
- OpenAI: git worktrees, cloud environments, code review, notifications and the Activity view
- OpenAI: models, other providers in
config.toml, pricing and limits - OpenAI: remote connections from your phone, scheduled tasks, the Windows app, the Linux preview
- OpenAI: Data controls
- Morphy: Chats and agents, Connect your AI, Freemium and Pro, Your files and privacy
