AWS announced two new microcredentials on September 29, and both are free. Securing Agent Identities Demonstrated and Databases for Agentic AI Demonstrated are live on AWS Skill Builder now.

If you haven't run into an AWS microcredential yet, here's the short version. It's a timed, hands-on lab in a live AWS environment. You get a broken or half-built setup, you fix it, and you pass or fail. AWS gives no hints along the way.

Securing Agent Identities

This one is aimed at cloud security engineers and solutions architects. The task is hardening identity and access across a fleet of AI agent workloads.

Expect workload identities, tag-based access control, JWT authorization, and the difference between SigV4 and JWT authentication. You also build audit trails. The lab runs on Bedrock AgentCore, Cognito, IAM, CloudTrail, CloudWatch, and Systems Manager.

This is the hands-on version of something we covered in September, when the SAP-C03 and DVA-C03 blueprints picked up AI security as tested material. Now there's a lab where you actually do it.

Can your champion sell internally without you? Because that's exactly what they have to do. Give buyers one workspace to evaluate, so your champion wins every meeting you're not a part of. Build Your First Aligned Room.

Databases for Agentic AI

This one is for database engineers, solutions architects, and AI/ML engineers. You build and troubleshoot the data layer that sits behind an agent.

That means Aurora PostgreSQL and DynamoDB configuration, pgvector setup, Bedrock Knowledge Bases, MCP servers, and query optimization. Lambda and Secrets Manager show up too.

I think pgvector and MCP servers are where most people will lose time on a first attempt. Plenty of folks have read about both and haven't set either up under a clock.

How it differs from a certification

AWS's own framing is that microcredentials prove practical ability, while certifications cover broader knowledge, and the two complement each other. That matches how these read to me. A cert tells an employer you know the material. A lab result shows you can do the work.

A few things AWS didn't say. The post doesn't list how long each lab runs, and it doesn't list prerequisites. I'd block off a quiet stretch instead of squeezing one in over lunch.

Also new this week

On September 30, AWS added 3 new Jam experiences. Each is a 1-day event. One covers spec-driven development with Kiro, one covers building agents on Bedrock AgentCore, and the advanced one covers multi-agent orchestration. They're booked through AWS classroom training, and the post doesn't give a price.

What to do

Studying for SAP-C03 or DVA-C03? Registration for both opens October 27. Try the Agent Identities lab in the meantime, since AI security is now on both blueprints and the lab costs nothing.

Work in security? Start with Securing Agent Identities.

Building agents on AWS? Start with Databases for Agentic AI.

Both are free, so the only real cost is an evening. Start from the AWS certification page if you want to see how the credentials fit together.

Keep Learning and Building.

If you're studying for Azure or AWS right now, Learn Cloud Academy makes two apps for exactly that. Learn Azure and Learn AWS each have 2,000+ practice questions, and every question links to the official documentation behind it. Miss one and you can read the real answer on the spot. Both are on iOS and Android, and more than 100,000 people have used them.

If you'd rather study in a browser, the same material is on learncloudacademy.com with Google Cloud added, and one price covers all 3 platforms.

If someone forwarded you this, the button below gets it in your inbox every Monday.