Bollard UI

AI agent safety

Keep a human in the loop when an agent can cause real change.

Show the action, scope and consequence before an agent sends a message, changes data or uses a sensitive value.

Try the guard rails

A small safety layer for one real flow.

  • Agent Action Approve

    Presents an agent action for explicit human approval.

    View component

  • AI Paste Guard

    Warns before sensitive or accidental content is pasted into an AI tool.

    View component

  • Undo Toast

    Makes reversible automation easy to correct immediately.

    View component

One command for this flow

Install the agent-safety pack.

Start with the related components, then edit the copied source in your own product.

npx shadcn@latest add @bollard/agent-pack

Add the Bollard namespace once

Merge this into your project's components.json. Then every Bollard install can use the short command.

{
  "registries": {
    "@bollard": "https://bollard-ui.vercel.app/r/{name}.json"
  }
}