SECURITY

Isolation for untrusted workloads.

AI agents can generate code, execute commands and interact with external systems. Workload Vault is designed to give those workloads a controlled execution boundary.

Execution isolation

Workloads execute in dedicated environments intended to separate their processes and files from the surrounding application.

Network controls

Define network access according to the requirements of your workload instead of assuming unrestricted connectivity.

Resource limits

Apply workload-level limits to help manage compute and runtime consumption.

Secrets

Keep sensitive credentials separate from generated source code and provide only the access a workload needs.

Logs & visibility

Make workload activity easier to inspect during development, testing and operations.

Lifecycle controls

Choose whether environments are temporary or persist for longer-running development workflows.

IMPORTANT

Security claims should match your implementation.

Workload Vault does not describe any system as “100% secure” or “hack-proof.” The exact isolation mechanism, infrastructure configuration and operational controls should be documented as the platform matures.