Preview the result
Try a rule against real facts and see what it would return. Nothing is recorded, so policy authors can refine an idea before anyone else sees it.
Policy decisions
An agent that can issue refunds needs a limit it cannot talk its way past. VeriProof checks each request against your active policy before the money moves, and hands back one of five answers instead of a yes or a no.
Allow or block is a gate. Real work needs more range than that, so VeriProof returns one of five outcomes: Allow, Warn, Transform, Escalate, or Deny. They match the verdicts in Microsoft's open Agent Control Specification. Each answer is one your compliance team can read without asking a developer what it meant.
For example Transform
A support agent requests $200, but the policy caps refunds at $150. VeriProof returns Transform with $150 as the corrected amount. The application issues the smaller refund.
If several rules match, VeriProof keeps every result and applies the strictest outcome, in this order: Deny, Escalate, Transform, Warn, Allow.
The check runs before the action, not after it. Your application sends the facts it has, VeriProof evaluates them against the current policy version, and the answer comes back while there is still a choice to make.
The action runs only once the answer lets it go ahead. A timeout or failed check releases nothing. Your application decides whether to retry or hand the work to a person.
Whatever happens next, the request, the rule, and the outcome stay linked as one record. That is what an auditor sees later. It is also the difference between having logs and having an answer.
Edge Gateway
Repeated checks add up in agentic workflows. Run eligible policies beside your application and send evidence to VeriProof in the background.
Your application key selects its rules automatically. Publish policies centrally; the gateway downloads and verifies updates.
Install the Edge GatewayExample response times
5× faster in this example.
Eligible gateway checks avoid the cloud round trip, reducing repeated network waits in high-frequency agent workflows.
Cloud API access remains available for shared controls and human review. Both routes keep decisions connected to the same application.
Try a rule against real facts and see what it would return. Nothing is recorded, so policy authors can refine an idea before anyone else sees it.
Shadow records what the rule would have decided without changing what the application does or opening review work. You find out how it behaves on live traffic first.
Authorized teams turn the reviewed rule on for the applications they choose. From then on, decisions apply at the point of action.
Change something that matters to the decision and the new version drops back to Shadow. Everything already recorded stays where it is.
A large payment needs finance to say yes, and the yes has to belong to that payment. VeriProof pauses the action, opens a review case with the facts and where each one came from, and rechecks identity and policy before anything resumes.
An approval cannot be reused for work that changed underneath it.
Monitoring tells you what happened. A policy check runs while there is still time to change or stop the action, which is the part monitoring cannot do.
Where the action runs. Your rules live in VeriProof, and the application that owns the workflow applies the answer through its SDK or integration, so enforcement sits next to the work rather than in a separate console.
You get the strictest outcome, and every matched result is kept with it. Reviewers can see which rules fired and why the final answer came out the way it did.
The customer portal provides SDK installation instructions. Explore the supported SDKs and release status to see the available language and framework options.
Bring the agents and decisions that matter to your team. We will walk through where each check goes, who reviews exceptions, and what the records show.