Inside theSoftware Factory
On deck

Figma logoFigma's Nimbus: AI sandboxes with 90% faster provisioning

Nimbus manages tens of thousands of AI sandboxes across multiple cloud providers, cutting provisioning latency 90% with warm pools.

What they built

Nimbus, a centralized platform for managing sandboxes at scale. It runs the secure execution environments that power Figma's AI features, including the design agent and Figma Make. The platform handles tens of thousands of sandboxes across multiple cloud providers, with a unified API that lets product teams request sandboxes as a configuration task rather than a provider integration.

How it works

Nimbus provisions sandboxes using a namespace configuration that specifies the container image, resources, and security controls. Product teams declare their intent, and the control plane handles provider selection, resource allocation, and lifecycle management. Warm pools of pre-provisioned sandboxes cut provisioning latency by 90% and absorb minutes of provider outages automatically. A sandboxd daemon inside each sandbox supervises workloads, runs health checks, and reports metrics. Security policies distinguish reads from writes, allowing features like web search while blocking unauthorized data exfiltration.

Key numbers

90%provisioning latency reduction from warm pools
99.95%provisioning availability (up from 99%)
Tens of thousandssandboxes under management

People

Pratik Agarwal Software Engineer, founding engineer on Nimbus
Jordan Rodgers Software Engineer, leads Sandbox Platform team

Related patterns

More On deck

We're keeping an eye on Figma. On deck companies aren't yet placed on a maturity level. With more detail on their whole setup, they'll get a full profile and a level.
Sources (1)
1
How Figma manages sandboxes at scale
Pratik Agarwal, Jordan Rodgers · 2026-11-04