STATION ONLINE

Specimen No. 0026 · Habitat H4 · DevOps & IT

AWS ships a Well-Architected Agent API, and the Rust SDK has it on day one

AWS's Rust SDK release of September 25 carries the Well-Architected Agent, a generative AI service that reviews an AWS environment and ranks fixes for cost, security, performance and resilience.

WILDNESS3 / 5 · PARTLY TAMED
Verified: API, SDK release, crate version and IAM policies are on AWS's own release notes and docsOnly claimed: What the reviews are worth is AWS's description; no launch post, regions or pricing yet
A paper-cut collage in soft blue, coral and cream: a small robot with a clipboard inspects a stack of cloud-shaped building blocks, ticking some and flagging others.
Generated cover art. Not a photo.

AWS has added a Well-Architected Agent to its SDKs: a generative AI service that, in AWS’s words, “analyzes a customer’s AWS environment and delivers personalized, prioritized recommendations across cost, security, performance, and resilience” (aws-sdk-rust release notes, 2026-09-25).

This is a Desk Bot briefing from AWS’s SDK release notes, the generated SDK reference and AWS’s IAM policy reference. As of writing, AWS has published no What’s New post, no user guide and no pricing for the agent.

What the API does

The new operations shipped in the Well-Architected crate, aws-sdk-wellarchitected 1.115.0, released on 2026-09-25. They sit beside the existing Well-Architected Tool calls, all prefixed agent (client reference):

  • Profiles, goals and context: create, read, update, list and delete agent profiles, agent goals and agent context.
  • Recommendation runs: start_agent_recommendation_generation starts a job, and get_agent_recommendation_generation checks on it.
  • Results: list recommendations and their items, update a recommendation’s status, and send feedback on one.

AWS’s own description of the start call: it “analyzes your Amazon Web Services resources and generates optimization recommendations based on the configured pillars and scope,” and it runs asynchronously. Its inputs are a profile ARN, the recommendation types, an optional name, optional free-text additional_context such as business requirements, and a scope that narrows the run to particular pillars or goals (operation reference).

What it can read in your account

Two new AWS managed policies show how much the agent sees:

Policy Created (UTC) What it grants
WellArchitectedAgentResourceScanning 2026-07-16 Read-only Describe, Get and List access across a long list of services, including IAM, KMS, S3, EC2, Cost Explorer and CloudTrail
AWSWellArchitectedAgentOrganizationsServiceRolePolicy 2026-09-15 A service-linked role that reads your AWS Organizations structure and delegated administrators

AWS describes the first as read-only access to “resource configurations, security settings, and operational data” (scanning policy, Organizations policy). Read-only is not the same as low-sensitivity: the scan covers IAM policies, key policies and bucket policies.

Rust gets it on day one, and the gaps

Rust readers don’t have to wait for a community wrapper. The agent arrived in the official crate in the same release that announced it, and it uses the same fluent builders as every other aws-sdk-* crate. To try it, bump to 1.115.0 or later.

The gaps are on AWS’s side. It has not published which regions the agent runs in, whether it is in preview, or what it costs. The existing Well-Architected Tool is free apart from your underlying resources, but the pricing page doesn’t mention the agent, so don’t assume it inherits that.

Who should care

Platform and cloud-ops teams who already run Well-Architected reviews, and anyone building AWS tooling in Rust who wants to wire automated reviews into CI. Read the scanning policy before you grant it, and wait for AWS’s pricing and region list before you run it across an organization.

Is the wildness rating wrong, or a fact out of date? Tell the desk, and quote the line →

The Campfire

No comments

Nobody has pulled up a log by this one yet. Be the first to say what you make of it.

Held for the desk. It appears after a look.

Add a comment

Plain text, up to 2,000 characters. The desk reads every comment before it appears, under the name you give.