Organizations building AI agents that autonomously write code, query databases, and interact with internal systems need secure execution environments, because without isolation one session’s tool calls can contaminate another session’s state, expose sensitive data across tenants, or allow untrusted code to reach production resources. Self-hosted sandboxes keep agent execution within your own AWS account, giving control over networking, secrets, and governance. AWS Lambda MicroVMs is a serverless compute environment providing general-purpose runtimes with the strong isolation of virtual machines and the rapid scaling of AWS Lambda. Powered by Firecracker virtualization, each MicroVM runs Amazon Linux with full OS access for up to 8 hours, and you launch, suspend, resume, and terminate MicroVMs programmatically. Three capabilities make it a strong fit for agent sandboxes: VM-level isolation per environment, with each MicroVM in its own Firecracker virtual machine; launch from snapshot, booting from a pre-captured memory and disk snapshot like Lambda SnapStart; and 4x vertical scaling without re-provisioning, scaling CPU and memory up to 4x its initial allocation, which can range from 0.25 vCPU/0.5 GB to 4 vCPU/8 GB, without terminating or re-creating the environment.