> ## Documentation Index
> Fetch the complete documentation index at: https://www.bolna.ai/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Implementing Guardrails for Bolna Voice AI Agents

> Discover how to set guardrails for Bolna Voice AI agents, ensuring safe, reliable, and compliant interactions tailored to your business needs.

## What are guardrails in Bolna Voice AI?

Guardrails are safety mechanisms that help you control and filter inappropriate content, prevent unwanted conversations, and ensure your Voice AI agents maintain professional boundaries during interactions.

<Frame caption="Guardrails on Bolna playground">
  <img src="https://mintcdn.com/bolna-54a2d4fe/mJ1zPF3eB4Dyupzj/images/guardrails_1.png?fit=max&auto=format&n=mJ1zPF3eB4Dyupzj&q=85&s=1fd9e4c71e8a651899ff14071cdbd4fe" alt="Bolna Voice AI Playground guardrails configuration interface showing how to add safety rules, content filters, and inappropriate language detection for AI voice agents" width="1933" height="1306" data-path="images/guardrails_1.png" />
</Frame>

<br />

## How to add guardrails to your agent

<Frame caption="Adding your own Guardrail">
  <img src="https://mintcdn.com/bolna-54a2d4fe/mJ1zPF3eB4Dyupzj/images/guardrails_2.png?fit=max&auto=format&n=mJ1zPF3eB4Dyupzj&q=85&s=02a2987ce299230e89a466eb6848ab60" alt="Step-by-step guardrail creation form in Bolna platform for adding custom safety rules including response triggers, threshold settings, and unwanted utterance detection for Voice AI conversations" width="1933" height="1306" data-path="images/guardrails_2.png" />
</Frame>

<br />

<Steps>
  <Step title="Name">
    Give your guardrail configuration an identifiable name to easily manage multiple guardrails across your agents.
  </Step>

  <Step title="Response">
    This is the action or message that gets triggered when any unwanted or inappropriate phrases or sentences are detected. Define what your agent should say or do when a guardrail is violated.
  </Step>

  <Step title="Threshold">
    This is the sensitivity level that determines when the system should react to unwanted words or phrases.<br /><br />

    <Note>If the threshold is set low, the system will react to even slightly inappropriate language. If it's set higher, only more severe cases will trigger a response. Adjust based on your use case requirements.</Note>
  </Step>

  <Step title="Utterances">
    These are the unwanted or inappropriate utterances which you want to guard against. List specific words, phrases, or topics that should trigger your guardrail.
  </Step>
</Steps>

## Why use guardrails?

Guardrails help you:

* **Maintain professionalism**: Prevent agents from engaging in inappropriate conversations
* **Ensure compliance**: Meet industry regulations and content standards
* **Protect brand reputation**: Control what your AI agents can discuss
* **Improve user experience**: Guide conversations toward productive outcomes

## Related features

Enhance your agent's behavior with other Bolna features:

* Configure [custom prompts](/agent-setup/agent-tab) to guide conversation flow
* Set up [context variables](/using-context) for personalized responses
* Implement [multi-agent workflows](/multi-agent-prompt) for complex scenarios


Built with [Mintlify](https://mintlify.com).