TLDRocket
Sign in

ICYMI: What landed for AI builders in September 2026

Amazon Web Services Prachi Mishra

AWS rolled out a batch of Bedrock, AgentCore, and Strands updates for AI builders. The big theme: more model choice, cheaper agents, and fresher enterprise data.

Based on reporting by Amazon Web Services, Prachi Mishra — read the original for the full story.

Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error

AWS spent September pushing a simple message: builders want choice, not a one-model future. The latest updates across Amazon Bedrock, Amazon Bedrock AgentCore, and Strands are built around that idea, with changes aimed at security, cost, speed, and the unglamorous work of keeping agents useful in real companies.

One of the bigger moves is Amazon Bedrock Managed Agents, powered by OpenAI, now in public preview. AWS says it lets teams build production-ready agents while keeping data inside AWS, reusing IAM permissions, and keeping audit trails through CloudTrail. Durable sessions and human approval workflows are part of the package too, which is a polite way of saying enterprise teams get more brakes and fewer surprises.

AgentCore also got a runtime update. The new version cuts cold start latency for serverless agents and improves memory management, so the pitch is lower cost and less infrastructure fuss. Sessions scale to zero when idle, run in hardware-isolated environments, and use pay-as-you-go pricing instead of pre-provisioned capacity. For teams trying to run agents all day without burning money on empty machines, that matters.

Strands is AWS’s open source answer for people who want portability. The new harness matches popular harnesses on accuracy while using 28 percent fewer tokens, and AWS says a production-ready agent can be started in a single line of Python or TypeScript. There is also Strands Decider 2B, a small open source decision model that picks between predefined options and returns answers in about 115ms locally.

The model lineup got wider too. OpenAI’s Astra, Sol, and Luna are generally available on Bedrock, alongside Claude Fable 5.1, Claude Opus 5.5, Claude Sonnet 5.5, Moonshot AI’s Kimi K3, and xAI’s Grok 4.6 and Grok 4.7. On the data side, Bedrock Managed Knowledge Base added automatic sync scheduling and new native connectors for ServiceNow, Confluence Data Center, Salesforce, and Zendesk. That is the less glamorous part of the stack, but it is also the part that decides whether an agent is smart or just confidently out of date.

My take — AI-written commentary, not fact-checked reporting

AWS is betting that enterprise AI buyers care less about model fandom and more about control, cost, and governance. That’s the right bet. The shiny demo is easy; the boring plumbing is where agent platforms either earn their keep or get quietly deleted by the security team.

Read more about this at: Amazon Web Services

Related stories

The daily briefing

Every AI story that matters, in your inbox by 8am.

TLDRocket reads all relevant sources, removes duplicate coverage, and summarises the day in two minutes. Follow companies and topics for alerts, or get the briefing in Slack. Free, no spam, unsubscribe anytime.