Amazon Q Developer Review (2026): AWS's AI Coding Assistant Up Close

Amazon Q Developer is the rebranded, expanded successor to CodeWhisperer, and by 2026 it has stopped being a pure autocomplete tool. It now ships inline completions, a chat panel, multi-step agents, a command-line agent, and code transformation jobs that rewrite whole projects. We spent time driving it inside VS Code and the CLI to see where the AWS pedigree actually pays off and where a generic editor like Cursor still pulls ahead.

The short version: if your day involves AWS — Lambda, IAM, CloudFormation, ECS — Q knows things no other assistant knows about your account. If you write plain application code with no cloud in sight, the gap between Q and the field narrows fast.

What you actually get

Q Developer installs as an extension for VS Code, the JetBrains IDEs, Visual Studio, and Eclipse, and as a standalone agent in the command line. Inside the editor you get the usual inline suggestions plus a chat sidebar that can read your open files. The two features that separate it from a stock code model are the agents and the transformations.

The feature-development agent (you trigger it with /dev in chat) takes a plain-English request, proposes a plan across multiple files, and generates a diff you review before applying. It is the same shape as Cursor’s agent or Claude Code, but wired into the AWS toolchain. The CLI agent does the same in a terminal and can run shell commands with your approval.

The transformation jobs are the part with no real equivalent elsewhere. Q can take a Java 8 or 11 codebase and upgrade it toward Java 17, fixing deprecated APIs and dependency versions as it goes, and it has a separate path for porting .NET Framework code to cross-platform .NET. These run as batch jobs against a whole module rather than a single suggestion.

Where the AWS integration earns its keep

This is the real differentiator. Q can read the context of your AWS account and answer grounded questions about it. Ask why an ECS task is failing and it can reference your actual service configuration; ask it to draft an IAM policy and it scopes to resources that exist rather than inventing ARNs. In the console it can help debug a Lambda function or explain a CloudFormation error against the stack you are looking at.

A generic assistant guesses at your infrastructure from whatever you paste into the prompt. Q queries it. For an engineer who spends the day in the AWS console and Terraform, that difference removes a lot of copy-paste and a lot of confidently-wrong answers about services that don’t exist in your region.

It also keeps the reference-tracking behavior CodeWhisperer introduced: when a suggestion closely matches public training code, Q flags it with the source and license so you can decide whether to accept it. If your legal team cares about provenance, that flag is worth more than it looks.

Pricing, and the honest tradeoff

Q Developer has a free tier that covers individual use with monthly limits on agent interactions and transformations, and a Pro tier at $19 per user per month that raises those limits, adds policy controls, and lets transformations run at larger scale. Pricing of this kind changes, so confirm the current limits on the AWS pricing page before you budget for a team.

The tradeoff to weigh: Q’s strength is coupling to AWS, which is also its constraint. Outside the AWS context, its raw code generation is competent but not obviously ahead of a model-agnostic editor that lets you pick Claude or GPT and tune the workflow to your repo. If you are multi-cloud or no-cloud, that coupling buys you less.

The practical move is not either/or. Several teams we have seen run Q for the AWS-shaped work — infrastructure questions, IAM, Java upgrades — and a general editor for everyday application code. The $19 seat pays for itself fastest on teams sitting on aging Java or .NET that they have been afraid to touch.

Who should pick it

Reach for Amazon Q Developer if you are already deep in AWS, if you maintain Java or .NET code that needs a version jump, or if license provenance on generated code matters to you. The account-aware answers and the transformation jobs are genuinely hard to replicate with a generic tool.

Look elsewhere first if your stack has little or no AWS surface, if you want to choose your own underlying model, or if you mostly want the fastest possible inline completions in a language Q supports unevenly. The tool is good; it is not trying to be the best general-purpose assistant, and you should not buy it as one.