Skip to main content
This quick start walks through a common workflow: select provider probes, create a profile, and deploy to an endpoint.

Step 1: Open Guardrails

  1. Sign in to Bud AI Foundry.
  2. In the left sidebar, open Guardrails.
  3. Click + Add Guardrail.
Image

Step 2: Select a Provider

  1. Choose a provider (for example, Bud Sentinel).
  2. Click Next to load available probes.
Image

Step 3: Choose Probes

  1. Use search and tags (for example: PII, Safety, Jailbreak).
  2. Select one or more probes relevant to your policy needs.
  3. Review probe details with See more if needed.
  4. Click Next.
Image

Step 4: Choose Deployment Type

Pick one:
  • Guardrail Endpoint: deploy as a standalone guardrail endpoint.
  • Add to Existing Deployment: attach to an existing model deployment.

Step 5: Configure Profile Settings

  1. Enter a guardrail profile name.
  2. Choose Guard Types (input/output as needed).
  3. Set a Severity Threshold.
  4. Optionally add a description.

Step 6: Select Project and Target

  1. Choose the destination project.
  2. If attaching to an existing deployment, pick the deployment/endpoint.
  3. Click Deploy.

Step 7: Verify Deployment

  1. Open the guardrail detail page.
  2. Check status and configuration in General, Probes, and Deployments tabs.
  3. Confirm the profile appears under the project’s Guardrails tab.

Governance Policy

Use this path when you want to govern an agentic workflow rather than only a model endpoint.

Steps:

Image
  1. Open Guardrails and select the Governance Policy tab.
  2. Click Add Policy.
  3. Add a name and description.
  4. Select a conflict strategy such as Deny / Escalate wins or Highest priority wins.
  5. Add policy checks for risks such as prompt injection, child safety, toxicity, spam, PII, or other available guardrail categories.
  6. Choose when each rule runs, such as before a run starts, before or after a model reply, or before or after a tool call.
  7. Set sensitivity from Lenient to Max.
  8. Choose the outcome: Allow, Redact, Notify, Escalate, or Deny.
  9. Add optional guidance, filter conditions, and priority.
  10. Save the policy, deploy it to a project, then attach it from the agent builder Governance section for the target agent version.

Next Steps

Concepts

Understand profiles, probes, and rule behavior.

Full Tutorial

Build a complete policy workflow.

Troubleshooting

Fix common setup and deployment issues.