OpenAI Codex goes mainstream

A coding assistant used to be a demo you tried once, like voice dictation in 2012. This week, OpenAI started treating Codex like standard developer infrastructure: the command-line version is open source, it runs locally in a terminal, and OpenAI says it can read, change, and run code in the directory you choose. (developers.openai.com) Codex is OpenAI’s software agent for programming, which means it does more than autocomplete a line. OpenAI’s product page says it is built for planning features, refactoring code, reviewing pull requests, and handling releases across the tools engineers already use. (openai.com) The new piece is the command line interface, which is the text window programmers use to control a computer with typed commands instead of buttons. OpenAI says Codex command line interface is built in Rust, runs on the developer’s own machine, and is open source instead of locked inside a closed app. (developers.openai.com) Open source changes the sales pitch because developers can inspect the code, modify it, and decide where it fits in their workflow. OpenAI is also bundling Codex into ChatGPT Plus, Pro, Business, Education, and Enterprise plans, which moves it from a separate experiment into the same subscription stack people already use. (developers.openai.com 1) (developers.openai.com 2) The pricing changed on April 2, 2026, and that is a bigger signal than it sounds. OpenAI’s help center says Codex for ChatGPT Business and new ChatGPT Enterprise plans moved from per-message billing to token-based billing aligned with application programming interface pricing, which is how developers already pay for core models. (help.openai.com) OpenAI made the same point publicly in a product post last week. The company said Codex now has pay-as-you-go pricing for ChatGPT Business and Enterprise teams, with no fixed monthly cost to get started, so a manager can roll it out the way they would roll out cloud computing instead of buying a seat for every engineer on day one. (openai.com) Then the usage numbers jumped. On April 8, 2026, Sam Altman said Codex had reached 3 million weekly users, and multiple reports said OpenAI reset usage limits to mark the milestone and plans to do that again at each additional million users until 10 million. (businesstoday.in) (technobezz.com) That reset matters because usage limits are the gate that turns a tool into a habit. If a developer hits the cap after a few refactors or test runs, Codex stays a sidekick; if OpenAI keeps widening access as usage grows, it becomes something teams can rely on during normal workdays. (businesstoday.in) (help.openai.com) OpenAI is also reshaping the menu around newer models instead of selling Codex as its own strange branch. The current Codex pricing and docs tie the product to the same model family and token logic used across OpenAI’s broader platform, which makes the coding assistant look less like a novelty bot and more like another front end for the company’s main models. (developers.openai.com) (help.openai.com) That is why this week feels different from the old “write me a function” era of coding assistants. OpenAI is putting Codex in the terminal, in team billing, in enterprise plans, and in the same operating rhythm as its flagship models, which is how products move from impressive trick to default tool. (developers.openai.com) (openai.com)