AWS DevOps Agent adds AI-powered release gates to speed up code deployment

The software development bottleneck has shifted. While AI agents now handle much of the code generation work, getting that code safely to production has become the constraint — and AWS is placing an intelligent gatekeeper at the merge queue.

AWS announced two new capabilities for its DevOps Agent on Wednesday at the AWS Summit in New York: a release readiness review and an autonomous release testing feature, both entering preview now at no extra cost.

What the release readiness review does#

Before code merges, the agent evaluates changes against production requirements in an AWS-managed isolated environment. It checks for cross-repository dependencies that could break other services, reviews access control changes against the AWS Well-Architected Framework, and validates compliance with team-defined standards (written in plain English, no custom tooling required).

The agent runs lightweight user journey tests to verify the build executes and passes basic functional checks. It then returns one of three verdicts: BLOCK, Proceed with Caution, or Safe to Release. Results appear in the DevOps Agent console, GitHub/GitLab pull requests, or directly in IDEs via the Kiro or Claude Code plugins.

Neha Goswami, AWS director of Agentic AI for Agentic DevOps, noted that this goes beyond static code analysis: "We have been able to find issues where the repo-level dependency was not being tracked by any of our existing solutions." In one early case, the agent blocked a change that human reviewers had already approved, catching a cross-repository dependency conflict they'd missed.

Autonomous release testing#

The second feature generates and runs change-specific test plans tailored to what the code actually does, rather than running a static test suite. These tests cover functional correctness, behavioral regressions, and integration scenarios, running in the customer's own production-like environment (unlike the readiness review, which uses AWS's isolated sandbox).

Each test run produces structured outputs: metrics, logs, traces, and an execution summary.

AWS's competitive angle#

Goswami emphasized that AWS's edge comes from testing these capabilities against Amazon's own massive engineering workflows before launch. "We have coded these agentic capabilities just the way we do this work," she said, pointing to the operational excellence practices baked into the agent's decisions about what to test and which guardrails to apply.