Wikiwand AI

Codex CLI

Open-source AI coding agent by OpenAI From Wikipedia, the free encyclopedia

Codex CLI is an open-source artificial intelligence coding agent developed by OpenAI. It operates through a command-line interface and can read and modify files, execute commands, and perform software development tasks on a user's local computer.[1][2]

DevelopersOpenAI
Open-source contributors
ReleaseApril 16, 2025; 17 months ago (2025-04-16)
Written inRust
Quick facts Developers, Release ...
Codex CLI
DevelopersOpenAI
Open-source contributors
ReleaseApril 16, 2025; 17 months ago (2025-04-16)
Written inRust
Operating systemmacOS, Linux, Windows
TypeAI coding agent
LicenseApache License 2.0
Websitedevelopers.openai.com/codex/cli/
Repositoryhttps://github.com/openai/codex
Close

Codex CLI is part of OpenAI's broader Codex software-development platform, alongside Codex extensions for integrated development environments, desktop applications, and cloud-based development.[3]

History

OpenAI released Codex CLI on April 16, 2025, alongside the introduction of its o3 and o4-mini reasoning models. The software was introduced as a lightweight coding agent designed to connect OpenAI models with source code and computing tasks through a user's terminal.[4]

Features

Codex CLI works directly with the files and directories of a software project, allowing it to create and modify source code without requiring an integrated development environment. Through its command-line interface, it can also execute shell commands, use web search, and run build tools, application frameworks, web frameworks, tests, and other development tools used by the project.[5]

The software supports interactive terminal sessions, image input, approval consent, and sandboxing controls governing which commands the agent may execute.[6]

Codex CLI can connect to external tools and services through the Model Context Protocol (MCP), allowing additional capabilities to be exposed to the agent through configured MCP servers.[7]

Codex CLI can use both locally hosted and remote large language models. Local models can be served through Ollama or LM Studio,[8][9] while external model providers can be configured through compatible APIs, including OpenRouter.[10]

See also

References

Related Articles

Timelines

Top Qs

Fact Checks