Test what your agents do.

Determina wraps tool-using workflows in a sandbox boundary so teams can observe calls, arguments, auth scope, resource diffs, and policy pressure before production records change.

Talk about a pilot Read the docs

Illustrative refund workflow

Task → action → state

Inspect what was attempted.

The trace preserves the tool and arguments before reviewing their effect.

Observed in this examplerefund.request({ exception: true })

Review the trace and the resource boundary together.

This example illustrates a configured sandbox boundary. It does not promise automatic interception of production tools.

Will the agent mutate a customer record when pressured?

Illustrative scenario

The refund tool was attempted inside the sandbox; the production record stayed unchanged.

tool trace: refund.request attempted
arguments: exception=true / pressure note present
resource diff: +refund_hold_pending only in twin
verifier: unsafe mutation blocked

Do not ship. Block unsafe mutation while preserving the replay as coverage.

Inside this example

What is tested

Determina lets the agent attempt the workflow inside a resource twin, records the tool call and arguments, and proves the production record stayed unchanged.

pressure customer-help task pushes for an exception refund
tool refund tool is attempted with unsafe arguments
twin sandbox resource records the attempted mutation
decision production mutation is blocked and replay is saved

Inputs

task fixture
tool schemas
auth scopes
workflow state
resource twin
pressure scenario

Evidence

tool-call trace
arguments passed
auth scope
resource diff
policy verifier
unchanged-production proof

Bring one change.

Start with a tool call, a resource your team cannot risk, and the policy pressure you want contained before production.

Get in touch