---
title: Bridge Policy and AgentVisor
slug: bridge-policy-and-agentvisor
createdAt: 2026-11-16T13:01:54.053Z
---

## Kyndryl Bridge Policy Service

:::ExpandableHeading{isOpen="true"}
### What is Bridge Policy Service?

The Bridge Policy Service is a comprehensive framework within Kyndryl Bridge designed to define, manage, and evaluate policies. It uses the Open Policy Agent (OPA) for decision-making. The policy framework evaluates decisions, while Bridge services enforce them, ensuring consistent governance and scalable control.
:::

:::ExpandableHeading{isOpen="true"}
### Why is a Bridge Policy Service important in  Kyndryl Bridge?

The Bridge Policy Service provides critical capabilities within Kyndryl Bridge :

- Decouples policy evaluation from application logic.
- Enables reusable decisions that are flexible and centrally managed.
- Ensures consistent governance and reduces risks.
- Streamlines policy management across security, compliance, and operational domains.
:::

:::ExpandableHeading{isOpen="true"}
### What is Open Policy Agent (OPA)?

An open-source policy engine that provides a way of declaratively writing Policies as Code (Rego) and then using those policies as part of a decision-making process. 
:::

:::ExpandableHeading{isOpen="true"}
### What is a Policy Bundle?

Policy Bundle is a package that includes the Rego (Policy as a Code) and Policy Configuration Data . OPA uses the policy bundle to evaluate decisions based on the input data received for policy evaluation from the bridge services or external services. Policy Bundle is stored in Git versioning system and imported to Bridge UI for policy evaluation.  
:::

:::ExpandableHeading{isOpen="true"}
### What is Rego?

&#x20;Rego is the programming language used to define policies (Policies as Code) in the Open Policy Agent (OPA). Rego enables the expression of rules and conditions that determine policy compliance and guide decision‑making within the Bridge Policy Service.
Bridge Policy Servic&#x65;**&#x20;rule example:** **Only Admins can create users in the application**.
:::

:::ExpandableHeading{isOpen="true"}
### What is Policy Configuration Data?

Policy Configuration Data refers to the structured data (typically JSON or YAML files) provided as input to policy evaluations. These configurations are external data that helps Rego scripts interpret and detrmine policy compliance and decision-making within the Policy service.&#x20;

For example, Admin User: 'Alice' is the configuration data that will be used for evaluating the policy rule in Rego "Only Admins can create users in the application"
:::

:::ExpandableHeading{isOpen="true"}
### What is Policy as Code?

Policy as Code refers to defining and managing policies in a machine‑readable, declarative language such as Rego for the Open Policy Agent (OPA). This approach enables automation, version control, integration with CI/CD pipelines, and more.
:::

:::ExpandableHeading{isOpen="true"}
### What are PDP and PEP?

- **PDP (Policy Decision Point):** Evaluates policies and returns decisions.
- **PEP (Policy Enforcement Point):** Enforces decisions based on PDP outputs.

The Bridge Policy Service functions as a PDP but not as a PEP. Enforcement of decisions is the responsibility of the client application that consumes the Bridge Policy Service.
:::

:::ExpandableHeading{isOpen="true"}
### What type of policies are supported?&#x20;

Policy Service supports the following policies:

- **Global Policies** managed by Kyndryl. They are centrally managed and reusable across accounts. Some may be mandatory while others are optional. 
- **Account/ Tenant Policies** created and managed for individual customer accounts.  They are specific to a customer account and allow customization of governance requirements to meet business, regulatory, or operational needs.  
:::

:::ExpandableHeading{isOpen="true"}
### How are policies evaluated?

A client application sends a policy evaluation request to Bridge Policy Service. The service identifies the applicable policy profile and policy bundles, enriches the request with additional context if required, evaluates the policies through OPA and returns a decision such as Allow, Deny, or Approval Required. The consuming application is responsible for enforcing the decision.  
:::

:::ExpandableHeading{isOpen="true"}
### Does Bridge Policy Service  provides auditability and logging?

Yes. Bridge Policy Service leverages the bridge service auditing and logging capability for auditing and logging various operations. 
:::

:::ExpandableHeading{isOpen="true"}
### Can policies be reused across services?

Yes. Bridge Policy Service is designed as a centralized decision service that can be used by multiple consumers including AgentVisor, automation workflows, applications, provisioning services, and compliance solutions.  
:::

:::ExpandableHeading{isOpen="true"}
### What are the user roles and permissions available for Bridge Policy Service&#x20;

The available user roles and permissions are:

- **Global Policies:&#x20;**&#x47;lobal Administrator for administering Global Policies 
- **Account/Tenant Policies:&#x20;**&#x50;olicy Administrator, Policy Viewer and Policy Evaluator for administering, viewing and evaluating Account Policies 
:::

# AgentVisor FAQs

:::ExpandableHeading
### What is AgentVisor?

AgentVisor is the runtime control and enforcement layer for Agentic AI. It intercepts agent actions, applies policy decisions received from Bridge Policy Service, and provides runtime visibility, auditability, and governance controls.  
:::

:::ExpandableHeading
### How do AgentVisor and Policy Service work together?

AgentVisor acts as the Policy Enforcement Point (PEP) and Bridge Policy Service acts as the Policy Decision Point (PDP). When an agent attempts an action, AgentVisor requests a policy decision from Policy Service and then enforces the returned outcome.  
:::

:::ExpandableHeading
### What policy decisions can be returned for an agent action?

Bridge Policy Service may return: 

- Allow 
- Deny
- Approval required
- Allow with additional conditions
- AgentVisor then enforces the returned decision
:::

:::ExpandableHeading{isOpen="true"}
### Can an agent use different policies for different use cases?

Yes. Different agents, workflows or business scenarios can be associated with different policies allowing governance controls to be tailored to specific operational requirements.  
:::

:::ExpandableHeading{isOpen="true"}
### Can users see why an agent action was denied?

Yes. Users can review agent execution logs to understand which policies were evaluated and the reasons why an action was allowed, denied, or interrupted.  
:::

:::ExpandableHeading{isOpen="true"}
### Can AgentVisor require human approval before an action is executed?

Yes. Bridge Policy Service can return approval-required outcomes for high-risk actions. AgentVisor can pause execution and route the action for review before proceeding.  
:::

:::ExpandableHeading{isOpen="true"}
### Are security policies required for agents?

Yes. Mandatory global security policies can be applied to ensure baseline governance and protection for agent actions across customer environments. Additional business policies may be activated as needed.  
:::

:::ExpandableHeading{isOpen="true"}
### Who owns security and business policies?

Typically, security policies are owned and managed centrally by security governance teams. Business policies are owned and managed by account teams or business stakeholders responsible for the use case.  
:::

:::ExpandableHeading{isOpen="true"}
### Can existing agents be governed by AgentVisor and Bridge Policy Service&#x20;

Yes. Existing agents can be onboarded into the AgentVisor governance model so that policy evaluation and enforcement can be applied without requiring a new agent implementation.  
:::

:::ExpandableHeading{isOpen="true"}
### What is the core governance principle used by AgentVisor and Bridge Policy Service ?

The recommended governance model follows a default deny approach, where actions are allowed only when explicitly permitted by the applicable policies. This helps reduce governance, security, and compliance risks.  
:::

:::ExpandableHeading{isOpen="true"}
### Do users need a separate license or entitlement to use AgentVisor and Bridge Policy Service ?

Access to AgentVisor and Bridge Policy Service is provided as part of the approved Bridge service and account onboarding model. Account teams should confirm any entitlement or access prerequisites before enabling agents. 
:::

:::ExpandableHeading{isOpen="true"}
### What should Delivery check before enabling an existing agent for an Account?

Delivery should ensure the agent is onboarded through the AgentVisor-enabled process for existing agents , the required policies are active in the account, and the agent's intended actions align with account governance requirements. No AgentVisor onboarding needed for new agents.
:::

:::ExpandableHeading{isOpen="true"}
### What happens if an agent requires policies that are not active in the account?

The account administrator must enable the required policies before making the agent available. Agent and policy configurations should remain aligned to avoid execution failures or unexpected policy decisions. 
:::

:::ExpandableHeading{isOpen="true"}
### What happens when a policy blocks or interrupts an agent action?

Users can review the execution logs of agents to understand what was allowed, denied, or interrupted. The appropriate follow-up may include policy updates, workflow adjustments, escalation, or user guidance. 
:::

:::ExpandableHeading{isOpen="true"}
### What is Principal Operation Resource Context (PORC)?

PORC is the standardized policy evaluation model used for AgentVisor: 

- **P&#x20;**- Principal: Who is performing or requesting the action? 
- **O&#x20;**- Operation: What action is being requested? 
- **R&#x20;**- Resource: What is the action being performed on? 
- **C** - Context: Under what conditions is the action being requested? 

To put it simple, PORC answers the question 'Who is trying to do what, to which thing, and under what conditions?' 
:::

:::ExpandableHeading{isOpen="true"}
### Why does AgentVisor use PORC?

AgentVisor needs a predictable structure for describing an agent's proposed action before policy can determine whether that action should proceed. 

PORC separates identity, requested action, target resource, and runtime/business conditions. This makes policies more consistent, reusable, explainable, and auditable across different agents. 
:::

