Amazon Restricts Claude Code Internally, Pushes Engineers Toward Its Own AI Tool Kiro

Amazon has moved to restrict its internal engineering teams from using Anthropic’s Claude Code, directing them instead to adopt Kiro, the company’s in-house AI coding assistant. Under the new policy, engineers working on production code or live projects will need formal approval to continue using Claude Code.

The Kiro Push

“We’ve seen dramatic improvements in efficiency and delivery speed with Kiro,” an Amazon spokesperson said. “We want our employees to take full advantage of these internal capabilities.”

Kiro is Amazon’s homegrown AI coding tool, built to integrate tightly with the company’s internal development workflows and infrastructure. The directive represents a clear signal that Amazon wants to consolidate its AI-assisted development under a single, company-controlled platform rather than relying on third-party tools.

Engineers Aren’t Happy

Not everyone inside Amazon is on board. Multiple reports indicate that engineers have raised concerns about Kiro’s performance relative to Claude Code. The complaints center on coding assistance quality — Claude Code has built a strong reputation for agentic coding capabilities, deep context understanding, and the ability to handle complex multi-file refactors.

Replacing a tool that developers have grown to rely on is never straightforward, and the friction is compounded when the replacement is perceived as technically inferior. Amazon will need to close the gap quickly or risk developer productivity losses during the transition period.

The Bedrock Paradox

The restriction creates an awkward contradiction in Amazon’s relationship with Anthropic. Through AWS Bedrock, Amazon actively sells Claude models to external customers as a premium AI offering. Amazon has also made substantial investments in Anthropic itself, reportedly totaling billions of dollars.

Yet internally, the company is now telling its own engineers to step away from Claude-powered tools. It’s a situation that underscores the tension between being an AI platform provider and an AI product company — Amazon wants to profit from Anthropic’s models externally while building competitive alternatives for internal use.

AI Sovereignty at Scale

Amazon’s move fits a broader pattern among major tech companies: the push for AI sovereignty. The strategic logic is straightforward — depending on an external AI provider for core development tooling introduces risk. If the provider changes pricing, performance, or terms, you’re exposed.

Google has been prioritizing DeepMind technology in its internal tools. Microsoft has positioned GitHub Copilot as its primary AI development platform. Now Amazon is making the same play with Kiro, even if the tool isn’t yet at parity with the competition.

The question is whether Amazon can iterate on Kiro fast enough to match what developers are giving up. Internal mandates can force adoption, but they can’t force satisfaction. If Kiro doesn’t improve rapidly, Amazon may find its engineers working around the policy rather than with it.