Skip to content

System Prompt

The system prompt is an instruction you give to the AI model that applies to the entire conversation. It sets the context, tone, and behaviour of the model before any user messages are exchanged.

What the system prompt does

The system prompt appears at the very beginning of the context that the model sees — before your first message. You can use it to:

  • Give the model a persona or role ("You are a helpful legal researcher")
  • Set boundaries on what the model should or should not discuss
  • Provide background information relevant to all your messages
  • Adjust the model's response style or format

Setting the system prompt

The system prompt editor is in the settings panel on the right side of the chat interface. Click in the text area and type your prompt, then click Save to apply it.

Settings panel showing the system prompt text area with a Save button below it

The default system prompt is:

You are a helpful assistant.

You can replace this with anything you like, or clear it entirely if you want the model to operate without any system-level instructions.

When changes take effect

System prompt changes take effect on the next message you send. Messages already in the conversation were generated with the previous prompt and are not affected.

System prompts and context

The system prompt counts toward the context window. A very long system prompt leaves less space for the conversation itself. Keep system prompts focused and concise unless you have a specific reason to include extensive detail.

See Context Window for more on how the system prompt is shown in the token usage breakdown.

Prompts per conversation

The system prompt is set per conversation. Different conversations can have different system prompts. If you start a new conversation, the system prompt resets to the default.