Testing your agent in the Playground

1. The Playground streaming chat

Open any agent in the editor, then click the run button. The Playground appears on the right: a live conversation window where you type prompts and watch the agent reply in real time. Each message streams word-by-word, so you see thinking happen. Use it to test the agent's behavior before publishing.

2. Context usage gauge

At the top of the Playground, a gauge shows how much of the model's context window you've used — e.g., 42% full. As the conversation grows (more messages, longer system prompts, bigger knowledge attachments), the gauge fills. If you approach the limit, consider archiving older messages or trimming the session memory to free space.

3. Copy generated artifacts

When the agent generates code, markdown, JSON, or other structured output, a copy button appears on that message. Click it to copy the artifact to your clipboard in one tap — no manual selection needed.

4. Voice dictation

A microphone icon in the input bar opens voice dictation. Speak your prompt; Groq Whisper-large-v3transcribes it to text. You're billed $0.0011 per minute of audio by duration. Press Send after dictation ends, or the transcript appears in the input field for you to edit.