MCP server that bridges Figma plugin data to AI tools without hitting API rate limits
This Figma MCP server demonstrates solid naming conventions and comprehensive descriptions, but significant gaps in parameter validation, schema completeness, and error handling prevent a higher score. All 11 tools have clear action verbs (get_, create_, save_) and descriptions between 150-400 characters, meeting baseline expectations. However, several parameters lack type constraints, descriptions are sometimes vague about formats/ranges, and error recovery guidance is absent. The schema definitions use Zod validators in source but the JSON Schema export (visible in the tools.ts snapshot) shows incomplete parameter documentation for critical tools. No tools include destructive/readonly/idempotent annotations. Output schemas are undocumented. The server performs well on naming (verb-first, clear distinctions between get_node/get_nodes/get_design_context) but falls short on schema rigor and error guidance expected of production tools.
Generate a design system rules file for your project. This creates instructions for AI coding agents to follow when implementing Figma designs, ensuring consistent component usage, design token mapping, and proper Figma-to-code workflow. Save the output to .github/copilot-instructions.md (VS Code) or .cursor/rules/figma-design-system.mdc (Cursor).
Get the design context for specific nodes by ID, or the current selection. Returns a structured representation with full layout (auto-layout/flex, sizing, padding, gap), typography (font family, weight, style, line height, letter spacing, text segments for mixed styles), visual styles (fills with variable bindings, strokes, effects, opacity, border radius), rotation, constraints, and component data. This is the primary tool for implementing Figma designs as code. Pass nodeIds to target specific nodes without needing to select them in Figma.
Get the current Figma page document tree with full serialized properties. Returns all nodes on the current page. For large pages, use get_metadata first to understand the structure, then get_design_context on specific nodes.
Get a structural overview of the current selection or page. Returns a lightweight tree of nodes with IDs, names, types, positions, and sizes. Use this tool first with large designs to understand the structure before calling get_design_context on specific nodes. Works with multiple selections or the whole page if nothing is selected.
Output schemas completely undocumented for all tools. LLMs cannot predict returned fields, preventing proper downstream tool chaining and result parsing.
No tool annotations (readonly, destructive, idempotent) present. LLMs cannot distinguish read-only tools from state-modifying ones. save_screenshot is WRITE but unmarked, risking unguided file system operations.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 60 | 2026-07-28+ | v2 |
Get a specific Figma node by ID with full serialized properties including layout, typography, visual styles, and component data. Must use colon format, e.g. '4029:12345', never use hyphens.
Get multiple Figma nodes by their IDs in a single call. Returns the full serialized data for each node including layout, typography (with per-segment breakdown for mixed styles), visual styles, rotation, constraints, and component data. Use colon-separated IDs, e.g. '4029:12345'.
Export a screenshot of the selected nodes or specific nodes by ID. Returns base64-encoded image data with dimensions. Use this alongside get_design_context for visual reference when implementing designs. PNG format preserves transparency (no white background). Use JPG only when transparency is not needed. SVG gives vector output with ID attributes.
Get the currently selected nodes in Figma with full serialized properties. Returns detailed layout, typography, and visual data for each selected node.
Get all local styles in the document including paint styles, text styles (with font family, weight, line height, letter spacing, text case, decoration), effect styles, and grid styles
Get all local variable definitions including variable collections, modes, and variable values. Variables are Figma's system for design tokens (colors, numbers, strings, booleans). Use this to extract color tokens, spacing values, and typography tokens used in the design.
Save exported screenshots of nodes to the filesystem. Takes an array of nodes with output paths and export parameters. Returns results for each node including file paths, dimensions, and success status.
Numeric parameters lack min/max constraints. 'depth' in get_metadata and get_design_context and 'scale' in get_screenshot have no documented bounds, inviting out-of-range values.
Array parameters have no length constraints. 'nodeIds' in get_design_context, get_nodes, and get_screenshot lack max count guidance. LLMs may pass hundreds of IDs in a single call, causing timeouts or API errors.
save_screenshot is a complex WRITE operation with no error recovery guidance or confirmation step. No documentation of what happens if outputPath is invalid, permissions denied, or disk full. No guidance on rollback/undo.
Enum values mentioned in descriptions but not formalized as schema constraints. create_design_system_rules describes 'react, vue, svelte, angular' but does not enforce them as an enum, allowing LLMs to pass invalid values.
save_screenshot's 'items' parameter is documented as an array but has no schema for array element structure. Nested fields (nodeId, outputPath, format, scale) lack individual descriptions.
No pagination or result limiting documented for tools that may return large data (get_document, get_design_context, get_nodes). Risk of exhausting context windows with verbose outputs.