Getting Started with the Playground
Follow these steps to create and test your prompts:- Choose a model - Select from available AI models
- Add System message - Define the AI’s role and behavior
- Add User message - Provide the input or query
- Configure parameters - Adjust creativity, response length, guardrails, and structured output
- Run to test - Execute your prompt and review the results
- Save the prompt - Store as a reusable template for future use

Managing Prompts
Saving Your Prompt in the Playground
Once you’re satisfied with your prompt’s performance, save it for future use. Click theSave
button and provide:
- ML Repo - Select the repository (see ML Repos)
- Name - Give your prompt a descriptive name
- Commit message - Add a brief description of changes


Loading Existing Prompts
The Playground allows you to work with previously saved prompts, enabling you to:- Build on previous work - Start from existing templates
- Collaborate with teammates - Share and iterate on prompts
- Continue where you left off - Resume prompt development


Advanced Playground Features
Input Variables - Dynamic Content
Input Variables - Dynamic Content
Create reusable prompts in the Playground with variables like 
Input variables configuration with dynamic placeholders.
{{customer_name}}
or {{product_type}}
.
You can then run these prompts by passing values. Please refer to Using Prompts for more details on running prompts with variables.
Guardrails - Configure Input and Output Guardrails
Guardrails - Configure Input and Output Guardrails
Configure guardrails in the Playground to ensure model responses stay appropriate and on-topic.Learn more: Guardrails and Security Overview
Guardrails configuration panel with available guardrail options.
Selected guardrail policies applied to the prompt.


Routing Config - Configure Load Balancing and Fallback Policies
Routing Config - Configure Load Balancing and Fallback Policies
Use the Playground to configure routing policies (load-balancing, fallback, retries) across models at the gateway.Learn more: Routing Configuration Overview
Routing configuration from Playground
Configured routing policies with fallback and load balancing settings.


Structured Output - Configure Response Format for LLM Responses
Structured Output - Configure Response Format for LLM Responses
Configure the Playground to get responses in specific formats such as JSON.
Structured output configuration with schema definition interface.
Generated JSON schema with defined properties and data types.


MCP Servers - Add Tools and MCP Servers
MCP Servers - Add Tools and MCP Servers
Use the Playground to extend AI capabilities with external tools and services.
MCP Servers configuration panel with available tools.
Configured MCP Servers with connected tools.

