We’re evaluating sandbox providers for a production AI agent that executes code on behalf of our users (similar to Manus or other agent-as-a-service products). Sprites caught our attention because the stateful/persistent nature would let us avoid rebuilding environments on every agent invocation.
However, the documentation and examples focus primarily on single-developer use cases (personal coding agents, individual sandboxes). We’re trying to understand if Sprites is a good fit for multi-tenant production workloads where:
-
We’d be spinning up Sprites programmatically via the API for third-party users
-
Each user session might need its own isolated Sprite
Is this a supported/intended use case? Are there any architectural recommendations or limits we should be aware of for this type of deployment?
