Amazon’s Kiro-first AI coding strategy is giving way to a broader toolkit

What changed inside Amazon

Business Insider reported that Jim Haughwout, Amazon’s vice president of Software Builder Experience, told employees in an internal note that Claude Code would be available company-wide immediately, with OpenAI Codex scheduled to follow on May 12 .

This does not appear to mean Amazon is letting developers connect freely to consumer-grade external tools. Reports say Claude Code and Codex will run in Amazon’s environment and be managed through Amazon Web Services, reducing the need for individual teams to build or maintain their own infrastructure; some reports also describe Amazon Bedrock as the hosting path .

That distinction matters. The question is no longer simply whether Amazon will allow third-party AI coding assistants. It is how the company can bring popular external tools into an approved, centrally managed access route.

Why the Kiro-first approach ran into resistance

Amazon had previously encouraged engineers to favor its own Kiro tool. Reuters reported on an internal memo in which Amazon suggested engineers avoid adopting third-party AI code-generation tools and use Kiro instead, while saying it had “no plans” to support additional third-party AI development tools at that time .

That stance met pushback from developers. Reports said around 1,500 Amazon engineers supported adopting Claude Code through internal forums, arguing that restrictions on preferred tools could slow workflows and hurt productivity . Business Insider also reported that Amazon had steered teams toward Kiro while requiring formal approval for Claude Code in production code or live products .

In other words, the pressure behind this change was not an abstract AI strategy debate. It came from working engineers who believed the tool policy was getting in the way of day-to-day development.

The practical reasons Amazon is broadening access

1. Developer preference became an operating issue

For a company with Amazon’s engineering scale, tool adoption is not just a procurement decision. If many developers think one approved tool slows them down, the result can be friction, workarounds, approval bottlenecks and weaker uptake.

That is the context for the reported support from roughly 1,500 engineers who wanted Claude Code access and pushed back against policies they associated with lower productivity . Opening Claude Code and Codex is a way to acknowledge that developer experience has to be part of the decision, not an afterthought.

2. Kiro is not necessarily finished — but it now has to compete

The available reporting does not provide a head-to-head benchmark proving that Kiro is technically weaker than Claude Code or Codex. The safer conclusion is that Amazon’s Kiro-first policy did not persuade all of its engineering teams, and that some developers wanted Claude Code added to the official toolkit .

That changes Kiro’s position. It can still benefit from being Amazon’s own tool, especially if it integrates well with internal systems. But it can no longer rely only on policy preference. It will have to prove its value in real development workflows alongside Claude Code and Codex.

3. Managed access is easier to enforce than blanket restriction

Amazon appears to be choosing control rather than laissez-faire access. Reports say Claude Code and Codex will run on Amazon’s infrastructure and be managed through AWS; some coverage says the tools will be available through Amazon Bedrock .

That gives Amazon a middle path. Engineers get more choice, while the company keeps a centralized route for infrastructure and deployment. Still, public reports do not disclose the full details of data isolation, logging, permissions, code-review rules or model-use policies, so it would be premature to say every governance question has been resolved .

4. Amazon’s AI partnerships make the decision less surprising

Business Insider framed the rollout against Amazon’s close relationships with Anthropic and OpenAI, describing Amazon as a partner of both AI companies . That context makes the internal adoption of Claude Code and Codex less like a retreat from Amazon’s own strategy and more like a broader ecosystem play.

For a cloud platform, supporting multiple AI tools can be useful both internally and commercially. The point is not that every team must use the same assistant, but that the company can provide an approved way to use several of them.

What this means for Kiro

The clearest reading is that Kiro is moving from privileged default to competitive option. Reports describe Amazon expanding beyond Kiro to include Claude Code and Codex; they do not describe Amazon shutting Kiro down .

That is still a meaningful demotion in practice. A tool that once benefited from a strong internal push now has to win developer trust in the same environment as high-profile external assistants.

The broader lesson for enterprise AI coding tools

Amazon’s reversal highlights a challenge many large companies are likely to face: one AI coding assistant may not fit every development team or workflow. In-house tools can be easier to govern, integrate and align with compliance requirements. But developers judge tools by whether they help them ship better code faster.

When a third-party assistant becomes popular inside the engineering organization, bringing it under central management may be more realistic than relying on restrictions and one-off approvals. That appears to be the logic behind Amazon’s new stance: keep Kiro in the mix, but do not sacrifice clearly expressed developer productivity needs to preserve a single-tool strategy .