Amazon rolls out Claude Code and Codex
- Amazon is giving all corporate employees access to Anthropic’s Claude Code now, with OpenAI’s Codex set to follow on May 12. - The key detail is where the tools run — inside Amazon Bedrock — so inference stays on Amazon’s own cloud infrastructure. - That matters because Claude Code had needed special approval, even as Amazon pushed its in-house Kiro across engineering teams.
AI coding tools are becoming normal company gear at Amazon — not side experiments for a few teams with special permission. The change is simple on paper: Claude Code is now available company-wide, and Codex is next. But the real story is what Amazon is standardizing. It is not just “use more AI.” It is a company-wide stack for how employees call these models, where the inference runs, and how the output gets governed inside the business. (businessinsider.com) ### What actually changed? Amazon told employees that Anthropic’s Claude Code is available immediately to all corporate staff, and that OpenAI’s Codex will be rolled out on May 12. The note came from Jim Haughwout, who leads Amazon’s software builder experience work. That turns a previously restricted setup into a default internal toolset. (businessinsider.com) ### Why is that a bigger deal than it sounds? Because Claude Code was not broadly available before. Earlier this year, Amazon guidance pushed teams toward Kiro — Amazon’s own coding assistant — and said Claude Code needed formal approval for production use. That created obvious f(businessinsider.com)icy was slowing them down. (africa.businessinsider.com) ### Why are Claude Code and Codex together? Amazon is basically admitting that one in-house tool is not enough. Developers want the best tool for the job, and right now that means mixing internal software with outside frontier models. Claude Code and Codex(africa.businessinsider.com) into the same workflow. (businessinsider.com) ### Why does Bedrock matter here? This is the load-bearing detail. Amazon is not just letting employees go sign into random third-party tools. The internal note said both products run through Bedrock, with inference staying on Amazon infrastructure. That gives Amazon tighter cont(businessinsider.com) enterprise plumbing. (aboutamazon.com) ### What happens to Kiro now? Kiro does not disappear. Amazon has said teams are still primarily using it, and one report said 83% of the company’s engineers had adopted it. So this is not Amazon abandoning its own assistant. It is more like Amazon conceding that internal standardization has to include external tools developers already want. (b17news.com) ### Why were employees pushing so hard? Because coding assistants are no longer a novelty perk. At a company this size, they are a productivity tool, and employees know when they are missing one. Internal criticism had built up around the idea that Amazon was limiting access to stron(b17news.com)arts looking less like governance and more like self-sabotage. (businessinsider.com) ### So what is Amazon really standardizing? Not the model. The process around the model. When multiple assistants are available, the competitive edge shifts upward — into review rules, testing pipelines, permissioning, logging, and how AI-written code gets checked before it reach(businessinsider.com)rth shipping. That is the part Amazon seems to be formalizing now. (aboutamazon.com) ### Bottom line? Amazon’s move says AI coding agents have crossed a line inside big companies. They are no longer experimental exceptions. They are becoming standard internal infrastructure — but with the company, not the model vendor, controlling the rails. (businessinsider.com)