Skip to main content
Node ReferencePrompt Node

Prompt Node

Author prompt text for downstream generation and processing nodes.

The Prompt node is a structured text input. You type instructions once, and downstream generation and processing nodes consume that text through a connection. It is the usual way to start a workflow with words rather than a model call.

Prompt is an input, not a generator
The Prompt node holds and emits text. It does not call a model. Connect it to a generation node such as Text Generation or Image Generation when you want the prompt to drive output.

What it stores

  • A single prompt string field holding the authored text.
  • The node starts empty. There is no model, provider, or parameter configuration on a Prompt node.
prompt node datajson
{  "type": "prompt",  "data": {    "prompt": "A wide shot of a lighthouse at dusk, cinematic lighting"  }}

Authoring a prompt

  1. Add a Prompt node

    Add a Prompt node from the canvas toolbar. A freshly spawned node shows a placeholder, “Describe what you want to create...”, with a brief shimmer to draw your attention to the input.

  2. Type your text

    The body is a multi-line text area. Edits are saved to the node's promptfield as you type. When you do not have edit access, the node shows “View only” and the input is read-only.

  3. Connect it downstream

    Drag from the Prompt node's output port to a generation node's prompt input. The footer shows a connection count once the node has at least one outgoing edge.

Character limit
Prompt text is capped at 1000 characters. The text area enforces this limit as you type, and input beyond the cap is rejected rather than truncated silently.

Ports and execution

The Prompt node runs locally — it does not make a network request. When the workflow executes, the node reads its text and emits it on the text-out port as a text value. Downstream nodes receive it on their prompt input (for example, prompt-in on Text Generation).

  • Whitespace-only text is treated as empty. If the resolved text is empty at execution time, the node fails with “Prompt node has no text content.”
  • The executor trims the text before emitting it.
  • If a value arrives on the node's own prompt input port, that connected value takes precedence over the stored field, followed by any expression input, then the stored prompt field.
Agent and API field fallback
The prompt executor reads data.prompt first. For compatibility with surfaces that write text to a data.text field instead (notably MCP tooling), a non-empty text field is used as a fallback when prompt is empty. Prefer writing prompt directly.

Related nodes

Use the Text node when you want a static text layer on the canvas for notes or labels rather than a runtime input. To learn how connections route values between nodes, see Edges and Prompt node architecture.

Was this page helpful?