You’re reading an issue of “The AI Economy,” my newsletter exploring the forces shaping the AI era—tracking how AI is rewriting business, work, technology, and culture. Subscribe to get expert insights and curated updates delivered straight to your inbox.
Amazon Web Services (AWS) is moving to eliminate one of the biggest time sinks in AI agent development. On Tuesday, the company introduced a managed agent harness within Amazon Bedrock AgentCore, a prebuilt infrastructure layer that handles orchestration, compute, tool connections, and session state, so developers can test agent logic without building backend plumbing first.
In addition, AWS is releasing an AgentCore Command Line Interface (CLI) to help build, deploy, and operate agents within a single terminal, along with a collection of prebuilt skills in AgentCore for Kiro, Claude Code, Codex, and Cursor.
Stop Worrying About the Plumbing
Before AI agents can work, developers have to spend time and resources building the backend plumbing, identifying the right framework, writing the orchestration code, connecting it to third-party tools and memory, and ensuring proper authentication. AWS’s managed agent harness provides all of this in a single package.
Because it’s powered by the Strands Agent framework, the open-source SDK that AWS released in 2025, it only requires three steps to get up and running. It starts by defining what your agent does, the model it runs on, what tools it calls, and the instructions it follows. Then, AgentCore does the rest.
For developers who want more control over their agent infrastructure than vertical platforms like Salesforce, ServiceNow, or SAP typically allow, building from scratch has been a daunting proposition. AWS is trying to change that with AgentCore’s new managed agent harness, which it says can get a working agent running in minutes without sacrificing flexibility.
Managed agent harness is now available in preview in four AWS regions: US West (Oregon), US East (N. Virginia), Asia Pacific (Sydney), and Europe (Frankfurt).
One Tool to Build and Control Your Agent
AWS is also collapsing the agent development lifecycle into a single command-line interface (CLI). Before AgentCore CLI, deploying an agent meant stepping entirely outside the development environment, configuring separate pipelines, standing up new environments, and rebuilding a process that bore little resemblance to how the agent was built in the first place. Now developers prototype, deploy, and operate from the same terminal without interruption.
Pre-Build AgentCore Skills For Vibe Coding
Finally, AWS is releasing a set of pre-built AgentCore skills for popular coding assistants, including Amazon’s Kiro, Anthropic’s Claude Code, OpenAI’s Codex, and Cursor. Rather than giving those tools generic API access, the skills encode curated knowledge of AgentCore best practices, so the suggestions developers receive reflect how the platform is actually meant to be used, not just what endpoints are available.
“Together, these capabilities create an unbroken path where you stay focused on agent logic and the infrastructure is handled by AgentCore,” the company writes in a blog post. “As your agent evolves, you add evaluations, memory, tool connections, and policy enforcement without rearchitecting. AgentCore is designed so the path from first test through deployment to ongoing operations stays on the same platform.”
Featured Image: AWS CEO Matt Garman speaks about updates to the company's Bedrock AgentCore platform at re:Invent 2025. Credit: Ken Yeung
You must be logged in to post a comment.