Codex CLI
Codex CLI is OpenAI's coding agent that runs locally on your computer, released under Apache-2.0. It installs with a script, npm or Homebrew, and pairs with a desktop app and extensions for VS Code, Cursor and Windsurf. It signs in with a ChatGPT plan or an OpenAI API key. The agent code is open; the models are OpenAI's hosted ones, so it does not run on your own machines end to end.
What does Codex CLI do?
"Codex CLI is a coding agent from OpenAI that runs locally on your computer." That is the README's own definition and its scope: the agent process is local, the model is OpenAI's. It runs in the terminal, as a desktop app started with codex app, and inside VS Code, Cursor and Windsurf through a separate IDE install. A separate cloud product, Codex Web at chatgpt.com/codex, is "the cloud-based agent from OpenAI."
Sign-in is the other half of the definition. The README recommends signing in with a ChatGPT account on the Plus, Pro, Business, Edu or Enterprise plans, and says an API key also works "but this requires additional setup." It names no model and no other provider, and mentions no local models. Standalone installers download from releases.openai.com with a fallback to GitHub Releases. File editing, command execution, MCP and sandboxing are covered on the documentation site rather than in the README, so they are not stated here.
Key facts
- Licence: Apache-2.0, with no added conditions, stated in the LICENSE file and the README.
- Runs locally as a terminal agent, as a desktop app (codex app), and in VS Code, Cursor and Windsurf through an IDE install.
- Sign-in: a ChatGPT Plus, Pro, Business, Edu or Enterprise plan, recommended, or an OpenAI API key with additional setup.
- Models: OpenAI's, not named in the README; no other provider and no local model is mentioned.
- Codex Web, at chatgpt.com/codex, is a separate cloud-hosted agent from OpenAI.
- Install: a curl script for Mac and Linux, a PowerShell installer for Windows, npm install -g @openai/codex, brew install --cask codex, or a prebuilt binary.
- Platforms: macOS Apple Silicon and x86_64, Linux x86_64 and arm64 including musl builds, Windows.
- Latest tagged release when read: rust-v0.155.0 on 2026-09-17; 124,929 stars on 2026-09-18.
What does it replace, and where does it fall short?
Codex CLI replaces Claude Code for a team on OpenAI models: the same shape, a local terminal agent with IDE extensions and a desktop app, with the agent's source under Apache-2.0 so you can read and change how it works. It appears on the open-source alternatives to Claude Code page as the pick for OpenAI models with open agent code.
Where it falls short of an open-source replacement: it needs an OpenAI account, and every model call goes to OpenAI, so nothing runs end to end on your own machines and no other provider or local model is offered. That is why the facts panel marks it as not running on your own machines. For a terminal agent that runs local models, Crush, Qwen Code or gptme.
How does Codex CLI run?
Install with the script at chatgpt.com/codex/install.sh on Mac or Linux, the PowerShell installer on Windows, npm install -g @openai/codex, or brew install --cask codex. Sign in with a ChatGPT plan, which the README recommends, or configure an OpenAI API key. The IDE extension for VS Code, Cursor and Windsurf installs separately, and codex app opens the desktop app. OpenAI's paid ChatGPT plans and Codex Web are the hosted side.
Who is Codex CLI for?
A developer already paying for ChatGPT or the OpenAI API who wants an inspectable, modifiable agent around those models. Anyone whose reason for going open source is to keep code off a vendor's servers should pick a different entry on the same page.
What limits does the README state?
From the README: using an API key rather than a ChatGPT plan requires additional setup. Nothing else is stated.
Questions people ask
Is Codex CLI open source?
The agent is: the repository is Apache-2.0 with no added conditions. The models are not; Codex CLI works only with OpenAI's hosted models through a ChatGPT plan or an API key, and the README mentions no other provider and no local model. This site labels the code open source and marks the tool as not running on your own machines.
Can Codex CLI run offline or with local models?
No, by the README. Sign-in requires a ChatGPT plan or an OpenAI API key, and every request goes to OpenAI's models. There is no provider setting and no local-model option described. Crush, Qwen Code and gptme are the terminal agents on this site whose READMEs document local models.
What does Codex CLI need to run?
A ChatGPT Plus, Pro, Business, Edu or Enterprise plan, which the README recommends for sign-in, or an OpenAI API key with extra setup. Then an installer for macOS, Linux or Windows, or npm or Homebrew. The IDE extension and the desktop app are optional additions.
How does Codex CLI compare with Claude Code?
They are the same shape from two vendors: a local terminal agent with IDE extensions and a desktop app, tied to that vendor's models. The difference is that Codex CLI's source is Apache-2.0 and readable. The open-source alternatives to Claude Code page places it beside OpenCode, Gemini CLI, Qwen Code, Crush, gptme and goose.
What is the difference between Codex CLI and Codex Web?
Codex CLI runs locally on your computer, in the terminal or through the desktop app and IDE extensions. Codex Web, at chatgpt.com/codex, is OpenAI's cloud-based agent that runs on OpenAI's infrastructure. The README describes both; only the CLI is the open-source repository.
Sources
- Codex CLI README and Apache-2.0 LICENSE: github.com/openai/codex, read 2026-09-17.
Compared with the others
On the open-source alternatives to Claude Code page, Codex CLI is the pick for openai models with open agent code. Apache-2.0 agent that runs locally, with a desktop app and IDE extensions, signed in with a ChatGPT plan or an API key.
Also on that page: OpenCode for a terminal agent with a planning mode, Qwen Code for switching between hosted and local models at runtime, Crush for the widest provider list with local models.
More coding assistants and agents
An MIT terminal coding agent with a full-access build agent and a read-only plan agent switched with the Tab key, installers for every platform, a beta desktop app, and the most-starred repository in this directory.
Google's Apache-2.0 terminal agent for Gemini models with a free tier of 1,000 requests a day on a personal Google account, built-in search grounding, file and shell tools, MCP support, checkpointing and a GitHub Action.
A high-performance code editor from the creators of Atom and Tree-sitter, GPL-3.0-or-later with Apache-2.0 components, with an agent panel, external agents over ACP, edit prediction, MCP tools, sandboxing, and local models through Ollama or llama.cpp.
An MIT self-hosted control centre, called Agent Canvas, that runs the OpenHands agent or Claude Code, Codex, Gemini and any ACP-compatible agent across local, Docker and cloud backends, with automations triggered by schedules and webhooks.
An Apache-2.0 coding agent for VS Code, JetBrains, the terminal and a desktop app that edits files as reviewable diffs, runs commands with your approval, works with any model provider including Ollama, and connects tools through MCP.
An Apache-2.0 general-purpose AI agent from the Agentic AI Foundation at the Linux Foundation, written in Rust, with a desktop app, a CLI and an API, 15+ model providers including Ollama, and 70+ extensions over MCP.
Added September 17, 2026. Every claim above comes from the project's README, LICENSE or model card, read on September 17, 2026, or from the GitHub API on the date shown in the panel. Found an error? Write to reveneau@licheo.com and it is fixed in the next weekly pass. Repository: github.com/openai/codex.
Running one of these inside your own environment, with your own data and your own security rules, is the kind of work Reveneau does. Read how a forward deployed engagement works.