Amazon Bedrock AgentCore Runtime: isolated microVMs for Claude Code, Codex, and Cursor

AWS has presented Amazon Bedrock AgentCore Runtime — a cloud execution environment for AI coding agents. Close your laptop and return to the task in the morning: the agent will continue working in the cloud independently, and results will be waiting in a persistent workspace.

What is Amazon Bedrock AgentCore Runtime

Amazon Bedrock AgentCore Runtime is AWS-managed infrastructure that provides each AI agent session with its own isolated microVM. The platform takes on the full lifecycle of a session — from startup to saving results. Files, cache, environment variables, and the entire working context are preserved between sessions and are not reset when the browser closes or the connection is lost.

Popular coding agents are supported: Claude Code, OpenAI Codex, Kiro, and Cursor. They can run in parallel within a single AWS account — each in its own microVM, without data, port, or secret overlap. AgentCore Runtime is built on top of Bedrock infrastructure and integrates with the rest of the platform's services, including access to tools through Gateway.

For teams that don't want to be locked into a single provider, support for agents from different ecosystems in a unified platform is particularly convenient.

Isolation and Security

The key principle of AgentCore Runtime is complete isolation between sessions. No agent sees the secrets, ports, or file system of another. In team development, where multiple engineers run different agents under one account, this is fundamental: even with parallel use, there is no risk of accidental token leakage or working directory conflicts. Access to external tools — APIs, databases, git repositories, enterprise services — is through AgentCore Gateway: a single entry point with authorization management and complete logging of every call. The agent operates with minimal privileges and requests only what is needed for the current task.

• microVM per session — isolation at the virtual machine level, not just a container
• Gateway for tools — one secure channel instead of direct credentials in the agent
• Persistent workspace — context, files, and cache live between session runs
• Built-in observability — tracing, logs, and metrics out of the box without additional configuration
• Parallel execution — multiple agents simultaneously without resource conflicts

The End of the Open Notebook Era

Before AgentCore, coding agents ran locally — right in the terminal or browser. Hours-long codebase analysis, major refactoring, test generation — all of this required keeping the machine on. Losing the connection or accidentally closing a tab meant losing progress.

This is especially painful for long tasks: a complete security scan of a large repository, documentation writing for hundreds of functions, or legacy code migration. Previously, such tasks were either broken into parts with context loss, or left on a server manually — with ad hoc scripts and no guarantees. AgentCore automates this "invisible" part.

Now you delegate a task to the agent, close your laptop, and go to dinner. The agent works in an isolated microVM on AWS's side — and in the morning the results are already in the workspace waiting for review. For teams, this means horizontal scaling without extra complexity: multiple developers run different agents simultaneously, while Gateway manages access centrally.

What This Means

AgentCore Runtime is a serious step by AWS toward transforming AI coding agents from desktop tools into full-fledged cloud infrastructure. For teams already using Claude Code, Codex, or Cursor, this is the next logical step: delegate not only tasks, but also their execution time. The pattern of "close the laptop — ready in the morning" becomes the standard, not the exception.