A strong fit
Teams connecting agents to private systems — you need execution close to internal APIs, databases or services without exposing those systems directly to the agent. Self-hosted AI platforms — you already operate your own model or agent stack and want the execution layer under the same operational control. Developers building agent products — you want a small execution primitive that can be embedded behind your own product and policies. Regulated or data-sensitive environments — you need to choose where code runs and where execution data is processed.May not be necessary when
- A developer-only local shell is acceptable for the task.
- A managed sandbox service already satisfies your infrastructure, data and governance requirements.
- The agent only calls narrow, pre-approved APIs and never needs general code execution.
- You do not want to operate execution infrastructure.