Post by Swami Sivasubramanian

🚀 Today we're making it easier for builders to go from idea to a running AI agent in minutes with three new features in Amazon Bedrock AgentCore: https://lnkd.in/gPxjMkff 🟠 Managed agent harness. This is the one I'm most excited about. Before today, teams had to build what we call the agent harness from scratch: the compute, tooling, memory, identity, and security that an agent needs to actually run. That meant days of setup before running a first real test. Now devs can declare their agent and run it in three API calls. Customers define their model, tools, and instructions, and AgentCore handles the rest. When organizations need custom orchestration or multi-agent coordination, devs can switch from config to code on the same platform without rearchitecting. Customers like VTEX, NTT Data, Parrot Analytics, and Reactiv are excited to take their agent ideas to production faster. 🟠 AgentCore CLI. Developers told us they wanted to stay in one workflow across the full agent lifecycle. The new CLI lets you prototype, deploy, and operate from the same terminal. Powered by infrastructure-as-code with CDK support and Terraform coming soon, so what devs test locally is exactly what runs in production. 🟠 New pre-built skills in AgentCore go beyond raw API access. They give coding agents curated, current knowledge of AgentCore best practices, so the suggestions you get reflect how the platform is actually meant to be used, not just what endpoints exist. Kiro already includes this today as a built-in Power, with support for Claude Code, Codex, and Cursor coming next week. On a platform that evolves quickly, having accurate context in your coding agent means fewer wrong turns from the very first line of code. Start building your next AI agent in just minutes on AgentCore 💻