MCP Tools
screenshot_node
Capture a PNG of a Figma node so the AI can see what was actually rendered.
Returns a PNG image of a node in the current Figma file. Call it after render_html (no node ID needed — it defaults to the root frame of the most recent render) or after edits, compare the image with what you intended, and fix differences with update_style, update_text, add_node or a new render_html. This closes the loop between generated HTML and the real Figma result.
Parameters
| Name | Type | Required | Description |
|---|---|---|---|
| nodeId | string | Optional | Figma node ID to capture. Defaults to the root frame of the most recent render_html. |
| maxWidth | integer | Optional | Width of the PNG in pixels (64–1600). Default: 1024. |
Example prompt
Try asking: Render the landing page, take a screenshot, and fix anything that doesn't match the design.
Tips
- Ask the AI to screenshot after every render so it can self-correct layout issues.
- Lower maxWidth for faster round-trips when only checking overall layout.
- Pass a nodeId to inspect a single component instead of the whole frame.