Screenshots and Metadata
When to Use
Two complementary tools for handling cases where get_design_context alone is insufficient: screenshots provide visual ground truth; metadata provides structure without weight for large frames.
Decision
| If you need... | Use... | Why |
|---|---|---|
| Visual reference alongside code generation | get_screenshot |
Layout fidelity check; reduces misinterpretation of structure |
| Identify node IDs in a large frame | get_metadata |
Sparse XML without styling data; cheap on token usage |
| Navigate a complex design before generating | get_metadata first |
Map the layer hierarchy, then target specific nodes |
| FigJam diagram content | get_figjam |
Returns metadata XML + screenshots for FigJam nodes |
Pattern
# Large frame workflow
"Get metadata for [figma-link] to understand its structure"
→ agent returns XML layer hierarchy with node IDs
"Now generate only the HeroSection node (ID: 123:456) in React"
→ agent calls get_design_context with specific node ID
# Visual accuracy workflow
"Get a screenshot of [figma-link] and then generate the HTML"
→ agent captures screenshot + calls get_design_context
→ uses visual reference to validate layout decisions
Common Mistakes
- Skipping
get_metadataon complex pages — going straight toget_design_contexton a 50-element page wastes tokens and often returns truncated results - Using
get_screenshotwithoutget_design_context— the screenshot alone gives the agent less structured information; use both - Forgetting
get_figjamexists for FigJam files — usingget_design_contexton a FigJam link will fail; the tool is FigJam-specific
See Also
- Code Connect Integration | FigJam and Diagram Tools
- Reference: https://developers.figma.com/docs/figma-mcp-server/tools-and-prompts/