Implementing Guardrails in Your AI Agent

Guardrails allow you to restrict your AI agent from responding to certain types of content. When a user’s message falls under a restricted category, the agent will reply with a predefined guardrail message instead of generating a normal response.

The platform already provides default guardrails such as hate speech, self-harm, and toxicity. You can also create custom guardrails based on your own requirements.

STEP 01

Go to agentic.botsify.com and select an agent.
 

STEP 02

From the left-side menu, click on Guardrails.
 

STEP 03

You will see default guardrails such as Hate Speech, Self-Harm, Toxicity, and others. If a user message falls under these categories, the agent will respond with the configured guardrail message instead of answering the query.
 

STEP 04

To create a custom guardrail, click on Add Rule.
 

 

STEP 05

Enter the name of the guardrail, for example Politics, add keywords such as politics or democracy, and add a custom response message that the agent should display when this guardrail is triggered.

STEP 06

Click on Create. The new guardrail card will be added to the list.

STEP 07

Click on Test Agent to open the conversation window.
 

 

STEP 08

Ask a question related to the restricted topic. The agent will respond with your custom guardrail message instead of answering the question.

 

You can create multiple guardrail rules, add more keywords, or define specific patterns to control the type of questions your agent is allowed to answer.