AWS Plans Continuum Integrations for Claude Code, Codex and Kiro

AWS says developers will be able to start an on-demand vulnerability scan from their coding environment. Potential findings will be sent to Continuum, which evaluates them against the customer's AWS configuration, identity and access policies, network topology and internet exposure. The service then validates suspected vulnerabilities in a sandbox before returning prioritized context to the coding assistant.

That sequence matters because a code-level warning does not always represent the same operational risk. A flaw in code that is not deployed should rank differently from one on an internet-facing path. AWS describes Continuum as an agent-team loop that can select different models for different parts of discovery, validation and remediation. Those capabilities and performance claims are vendor-reported and will need independent evaluation as the preview expands.

AWS outlined two intended modes. Existing applications can use Continuum across the deployed environment, while greenfield work can use a plugin in Claude Code, Codex or Kiro to receive security-validated suggestions during development. SiliconANGLE separately reported that the service can build a working exploit in an isolated sandbox to reduce false positives before recommending remediation.

The announcement moves security feedback closer to the point where agent-generated code is written, but it does not remove the need for review. Teams evaluating the integrations should examine how repository access, AWS permissions, sandbox evidence, approval gates and audit logs are scoped before letting an agent propose or apply a fix.

The most important availability detail is also the easiest to miss: the coding-tool integrations have been announced, not shipped. For now, teams can request access to the Continuum preview and assess whether its context and validation model fits their development and security controls.