Run Codex CLI in MXterm: install, first session and a split pane for tests

The short version

1. Install Codex from OpenAI's instructions.

2. Open a new tab in MXterm, choose your project folder and add one Codex pane from the agent list.

3. Give Codex a task, then split the pane for a shell.

Install Codex

OpenAI's Codex CLI page says: "Install the Codex CLI with the standalone installer for macOS and Linux." It also has tabs for Windows, npm and Homebrew, so the standalone installer is not the only route (learn.chatgpt.com/docs/codex/cli, read 31 October 2026).

The same page says: "Open a project directory and run codex. The first time you run Codex, choose Sign in with ChatGPT or another available sign-in method." OpenAI's README says "Codex CLI is a coding agent from OpenAI that runs locally on your computer" (github.com/openai/codex, read 31 October 2026).

Commands change, so check OpenAI's page on the day you install. MXterm launches the agents you have installed; it does not install them.

Step 1: add a Codex pane from the agent list

Open a new tab in MXterm and choose your project folder. The default agent list leads with Claude Code, Codex, Grok and Gemini CLI, then a plain Terminal row and OMP, and lists every other agent after them. In that default list Codex is one of the first rows.

Each installed agent has plus and minus buttons that set how many panes of it the tab opens; press Create workspace to open the tab. An agent that is not installed shows Not installed and an Install button, which opens the install instructions or shows the command.

Step 2: give it a task

Type the task as you would in any terminal; the pane is a real shell.

Codex asks you to sign in the first time you run it. MXterm does not sign you in.

Step 3: a shell beside it

Use the split buttons in the pane header (Split Right and Split Down) to put a plain shell next to Codex for your tests. Split panes run real shells.

Run your test command in the shell pane, and keep Codex in the pane beside it.

What MXterm adds for Codex

See which agent is working, waiting or done: amber means working, red means blocked and green means done.

When Codex stops to ask for permission, the question card on your desktop shows every answer the agent offers. Chat view shows the conversation as a chat, and you switch between Chat and Terminal in the pane header; it is available for Codex.

To see how much a session used, the token chip in the title bar shows token use at a glance, computed on your Mac.

Codex next to Claude Code

Codex and Claude Code can run in separate panes of the same workspace, and each agent can work on its own branch in its own worktree. That reduces collisions; it does not stop two agents editing the same file.

The guide to running both, Claude Code in MXterm and isolating agents with git worktrees have the details. To keep agents in one room, see the Control Room. Our list of AI coding agents in 2026 gives the vendors' own descriptions of each one.

When a plain terminal is enough

For a Codex task, a plain terminal is enough once Codex is installed and you are signed in.

MXterm helps when agents or tasks run in parallel and you want their state, their branches and their questions in one window. To try it, use the download page.