Skip to main content
Tensorlake provides isolated MicroVM sandboxes that boot in hundreds of milliseconds, with memory and filesystem preserved across suspend/resume. Get setup in a few minutes, and start a sandbox in a few seconds. Sandboxes can be used to run agent harnesses, run tool calls or even as VMs for running coding agents, builds and IDEs.

​
How it works

Sandboxes are created on-demand via API calls, and they are MicroVMs backed by Firecracker and CloudHypervisor. You can specify images and resources when creating them. The default image, tensorlake/ubuntu-minimal, starts up in a few hundred milliseconds, while tensorlake/ubuntu-systemd has a full init system and more tools and takes around 1 second to boot. 
  tl sbx create

​
Start Using Sandboxes

Quickstart

Install the SDK and run your first sandbox.

Lifecycle

The mental model behind ephemeral, named, suspend, and snapshot.

Sandbox Images

Use and customize sandbox images for your use case.

Sandbox Persistence

How to persist sandbox state across runs with suspend and snapshots.

​
Trust and support

Tensorlake is HIPAA and SOC 2 Type II compliant, supports EU data residency, and offers zero data retention.

Slack

Chat with our engineers.

Email

support@tensorlake.ai

Newsletter

Use cases and product updates.