Playground

What is the Playground?

The Playground lets you test prompts, models, settings, and tools before you add them to a Flow or Agent. Run a conversation in the dashboard before you create a saved Flow or Agent.

How the Playground works

Configure a model and system prompt, add optional variables and Tools, then send messages in the chat panel. Each message uses your selected configuration and the context from earlier messages.

The model list depends on the models enabled for your account and the provider keys available to you. For more information about key options, see Platform keys vs. bring your own key (BYOK).

Use the Playground to:

  • Refine prompts: Change wording, instructions, and output formats.
  • Compare models: Send the same prompt to different models and compare the responses.
  • Tune settings: Adjust temperature, response length, response format, and reasoning.
  • Test tools: Select Tools and review tool calls in the conversation. For more information, see What are Tools?.
  • Prepare a Flow or Agent: Test a configuration before you save it as a Flow or Agent.

Example

Suppose you are Dana, and you are building a customer support chatbot for refund requests. Test the configuration in the Playground before you add it to a Flow:

  1. Select gpt-5.6-terra and write a system prompt that states your refund policy.
  2. Test edge cases such as angry customers, missing order numbers, and requests outside the return window.
  3. Select claude-sonnet-5 and send the same scenarios.
  4. Set Temperature to 0.3 to reduce variation between replies.
  5. After the prompt produces acceptable replies, click Save as Flow.

The Save as Flow dialog creates a Flow with one Prompt step. It copies the Model, System Prompt, Temperature, Max Tokens, and selected tools into that step.

Open the Playground

To open the Playground, click Playground in the sidebar. The page selects your account’s default enabled model after the model catalog loads.

Use the Playground interface

The Playground has a configuration panel and a chat panel. On desktop, resize the panels with the divider. On mobile, open the chat panel from the header.

Select a model

In the configuration panel, select a model from the Model field. The list includes enabled models that your account can use. Select a model before you add Tools.

Enter a system prompt

In the System Prompt field, enter the instructions that guide the model’s response. Include variables such as {{CUSTOMER_NAME}} and {{ORDER_ID}} when you want to test different inputs. These placeholders represent values that you enter before the run.

The Playground detects each variable and displays a field where you can enter its value. It substitutes those values into the prompt before the run.

Adjust advanced settings

To adjust model behavior, click Advanced settings. The Advanced Settings panel provides the following options:

  • Response Format: Use the model default, Markdown, JSON, HTML, or XML.
  • Temperature: Set a value from 0 to 1. Lower values reduce variation, and higher values increase variation.
  • Max Tokens: Set a response length from 1 to 128,000, or use the model default.
  • Reasoning: Use the model default, enable reasoning, or disable reasoning when the model supports the setting.

Continue a conversation

The chat panel keeps the messages in the current conversation. Send another message to test how the model responds with the preceding context.

Run a test

To run a test, follow these steps:

  1. In the configuration panel, select a model from the Model field.
  2. In the System Prompt field, enter your prompt.
  3. Optional: Add variables in the {{VARIABLE_NAME}} format. Enter a value for each variable that the Playground detects.
  4. Optional: Open Advanced Settings and adjust the model settings.
  5. Optional: In the Tools field, select the tools that you want to test.
  6. In the chat panel, enter a message and send it.
  7. Review the response in the chat panel.

Send additional messages to test a multi-turn conversation. The Playground includes the earlier messages as context for each new message.

Save your configuration

After you finish testing, choose a save action:

  • Click Save as Flow to create a Flow with one Prompt step. The saved step includes the Model, System Prompt, Temperature, Max Tokens, and selected tools.
  • Click Save as Agent to create an Agent with the Model, System Prompt, Temperature, and selected tools.

After you choose a save action, enter a name in the dialog and click Create flow or Create agent. The save action does not copy the Playground’s Response Format, Reasoning selection, or variable values.

Next steps