Local-first Figma Desktop MCP bridge for Codex and compatible MCP clients.
Strong tool naming with clear action verbs (figma_get_*, figma_status). All 7 tools have descriptions (avg 140 chars, within 10-1024 range). Input schemas are complete with proper JSON Schema types, constraints (min/max), and defaults. However, parameter descriptions are sparse or missing entirely, most params lack explanatory text beyond the name. Output schemas are undocumented; LLMs cannot predict response structure. No error handling guidance visible. All tools are read-only (low risk), but descriptions could be more LLM-optimized per pattern:tool-description baseline (50-200 chars ideal).
Build an implementation-ready context pack for the current selection or a Figma node, including bounded structure, tokens, components, variables, styles, assets, and an optional screenshot.
Read metadata for the Figma file currently open in Figma Desktop.
Read one Figma node by ID, including its bounded child tree.
Read the top-level nodes on a Figma page. Defaults to the current page.
Export the current selection or a Figma node as a PNG image for visual comparison.
Read the current Figma selection, including layout, text, fills, and a bounded child tree.
Parameter descriptions missing or minimal. 'pageId', 'depth', 'maxChildren', 'maxNodes', 'scale', 'includeScreenshot' lack explanatory text. LLMs cannot infer parameter semantics from names alone.
Output schemas not documented. Tools return complex nested structures (bounded child trees, tokens, components, variables, styles, assets, screenshots) but LLMs have no schema to predict response fields or plan downstream calls.
No error handling guidance. Descriptions do not explain failure modes (e.g., 'plugin not connected', 'invalid nodeId', 'timeout') or recovery steps. Agents cannot self-correct on errors.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 66 | 2026-07-28+ | v2 |
Check whether the local Figma Desktop plugin is connected and report the active file/page.
Parameter constraints not explained in descriptions. 'depth' ranges 0-6, 'maxChildren' 1-300, 'scale' 0.5-4, but descriptions do not state these bounds or their semantic meaning (e.g., 'depth: tree traversal depth in levels').
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). All tools are read-only and idempotent, but this is not declared in the tool definitions, forcing LLMs to infer safety properties.