figma-console-mcp
v1.46.0
Published
The most comprehensive MCP server for Figma — design tokens, variables, components, write tools, version history diff, accessibility audits, FigJam, Slides, and more. Local (WebSocket Desktop Bridge plugin) and Cloudflare Workers (paired + remote) modes.
Maintainers
Readme
Figma Console MCP Server
Your design system as an API. A design-system-focused Model Context Protocol server for Figma. It works in both directions (Figma ⇄ code), writes to Figma, and runs deterministic checks for design-code parity, accessibility, and design-system health. It hands the AI exact tokens, variants, bindings, and states as structured data rather than prescribing a framework or house style, so generated code follows your team's own stack and conventions.
🆕 Shaders, grid and auto layout, OKLCH and P3 color, text wrap and variable fonts (latest v1.46.0): apply shaders and set their properties by name (
figma_apply_shader,figma_list_shaders), set auto layout or grid layout in one call with CSS-style grid tracks (figma_set_layout), use OKLCH, Display P3 and other CSS colors on fills and strokes, and set text wrap (balance, pretty) and variable-font axes withfigma_set_text(v1.46.0). v1.45 added Motion: animate layers with presets, custom animation styles or keyframes bound to Timing and Easing tokens, export video, and find code layers. See what's new →
What is this?
Figma Console MCP connects AI assistants (like Claude) to Figma, enabling:
- 🎨 Design system extraction - Pull variables, components, and styles
- 🧬 Codebase → design system extraction - Scan a production codebase, mine its de-facto styling into DTCG tokens with per-token provenance, scaffold a design-system package with a wired Storybook workshop, and gate it with deterministic fidelity evals — then round-trip the tokens into Figma variables (Local Mode)
- 🔁 Bidirectional token sync - Export Figma variables to DTCG JSON (legacy or 2025.10 dialect) + 9 more formats; push code-side edits back to Figma with full apply — creates, renames, alias re-targeting, and replace-gated deletes. Replaces Style Dictionary and Tokens Studio's export pipeline.
- 📸 Visual debugging - Take screenshots for context
- ✏️ Design creation - Create UI components, frames, and layouts directly in Figma — including one-call component-set creation from a variant axes matrix
- 🔧 Variable management - Create, update, rename, and delete design tokens
- 🕰 Version history & time-series awareness - List versions, diff snapshots, generate markdown changelogs, trace property/variant introduction via binary-search blame, and fold per-component design + git history into generated component docs
- ⚡ Real-time monitoring - Watch console logs from the Desktop Bridge plugin
- 📌 FigJam boards - Create stickies, flowcharts, tables, and code blocks on collaborative boards
- 🎞️ Slides presentations - Build and manage Figma Slides decks programmatically
- ♿ Accessibility scanning - 14 WCAG design checks with conformance level tagging, component scorecards, axe-core code scanning, design-to-code parity
- 🔍 Design-code parity & component docs - Deterministic comparison of a Figma component with its code (visual, spacing, typography, tokens, component API, accessibility), and generated component documentation with per-variant tokens and an optional design + git changelog
- 📋 Design-system health audits - Lighthouse-style scored audit across naming, tokens, component metadata, accessibility, consistency, and coverage — available as a plain tool for any MCP client (with per-finding auto-fix guidance) and as a visual dashboard app
- 🛡 Cross-MCP identity - Every tool response carries
_mcp: "figma-console-mcp"and errors are prefixed[figma-console-mcp]so attribution stays unambiguous in agents running multiple Figma MCPs - ☁️ Cloud Write Relay - Web AI clients (Claude.ai, v0, Replit) can design in Figma via cloud pairing
- 🔄 Four ways to connect - Remote SSE, Cloud Mode, NPX, or Local Git
⚡ Quick Start
Choose Your Setup
First, decide what you want to do:
| I want to... | Setup Method | Time | |--------------|--------------|------| | Create and modify designs with AI | NPX Setup (Recommended) | ~10 min | | Design from the web (Claude.ai, v0, Replit, Lovable) | Cloud Mode | ~5 min | | Contribute to the project | Local Git Setup | ~15 min | | Just explore my design data (read-only) | Remote SSE | ~2 min |
⚠️ Important: Capability Differences
| Capability | NPX / Local Git | Cloud Mode | Remote SSE |
|------------|-----------------|------------|------------|
| Read design data | ✅ | ✅ | ✅ |
| Create components & frames | ✅ | ✅ | ❌ |
| Edit existing designs | ✅ | ✅ | ❌ |
| Manage design tokens/variables | ✅ | ✅ | ❌ |
| FigJam boards (stickies, flowcharts) | ✅ | ✅ | ❌ |
| Real-time monitoring (console, selection) | ✅ | ❌ | ❌ |
| Codebase → design system extraction (figma_ds_*) | ✅ | ❌ | ❌ |
| Desktop Bridge plugin | ✅ | ✅ | ❌ |
| Requires Node.js | Yes | No | No |
| Total tools available | 130 | 105 after pairing | Read-only subset |
Bottom line: Remote SSE is read-only until you pair the plugin. Cloud Mode unlocks write access (105 tools) from web AI clients without Node.js. NPX/Local Git gives the full 130 tools with real-time monitoring.
🚀 NPX Setup (Recommended)
Best for: Designers who want full AI-assisted design capabilities.
What you get: All 130 tools including design creation, variable management, and component instantiation.
Prerequisites
- [ ] Node.js 18+ — Check with
node --version(Download) - [ ] Figma Desktop installed (not just the web app)
- [ ] An MCP client (Claude Code, Cursor, Windsurf, Claude Desktop, etc.)
Step 1: Get Your Figma Token
- Go to Manage personal access tokens in Figma Help
- Follow the steps to create a new personal access token
- Enter description:
Figma Console MCP - Set scopes (all Read unless noted): File content, File versions, Library content, Library assets, Comments (Read and write), Current user, and Variables (Enterprise only, optional). See why each is needed
- Copy the token — you won't see it again! (starts with
figd_)
Step 2: Configure Your MCP Client
Claude Code (CLI):
claude mcp add figma-console -s user -e FIGMA_ACCESS_TOKEN=figd_YOUR_TOKEN_HERE -e ENABLE_MCP_APPS=true -- npx -y figma-console-mcp@latestOpenAI Codex (CLI):
codex mcp add figma-console --env FIGMA_ACCESS_TOKEN=figd_YOUR_TOKEN_HERE --env ENABLE_MCP_APPS=true -- npx -y figma-console-mcp@latestThen set startup_timeout_sec = 60 and tool_timeout_sec = 300 for it in ~/.codex/config.toml (Codex's 10 s / 60 s defaults are too short). The Codex app and IDE extension share this config. Full Codex setup →
Gemini CLI:
gemini mcp add -s user -e FIGMA_ACCESS_TOKEN=figd_YOUR_TOKEN_HERE -e ENABLE_MCP_APPS=true figma-console npx -y figma-console-mcp@latestGoogle Antigravity (IDE, CLI, app): add the JSON below to ~/.gemini/config/mcp_config.json, plus "timeoutSeconds": 300 in the server entry. Full Antigravity setup →
Cursor / Windsurf / Claude Desktop:
Add to your MCP config file (see Where to find your config file below):
{
"mcpServers": {
"figma-console": {
"command": "npx",
"args": ["-y", "figma-console-mcp@latest"],
"env": {
"FIGMA_ACCESS_TOKEN": "figd_YOUR_TOKEN_HERE",
"ENABLE_MCP_APPS": "true"
}
}
}
}📂 Where to Find Your Config File
If you're not sure where to put the JSON configuration above, here's where each app stores its MCP config:
| App | macOS | Windows |
|-----|-------|---------|
| Claude Desktop | ~/Library/Application Support/Claude/claude_desktop_config.json | %APPDATA%\Claude\claude_desktop_config.json |
| Claude Code (CLI) | ~/.claude.json | %USERPROFILE%\.claude.json |
| Cursor | ~/.cursor/mcp.json | %USERPROFILE%\.cursor\mcp.json |
| Windsurf | ~/.codeium/windsurf/mcp_config.json | %USERPROFILE%\.codeium\windsurf\mcp_config.json |
| OpenAI Codex (app, CLI, IDE) | ~/.codex/config.toml (TOML, see format) | %USERPROFILE%\.codex\config.toml |
| Gemini CLI | ~/.gemini/settings.json | %USERPROFILE%\.gemini\settings.json |
| Google Antigravity (IDE, CLI, app) | ~/.gemini/config/mcp_config.json | %USERPROFILE%\.gemini\config\mcp_config.json |
Tip for designers: The
~symbol means your home folder. On macOS, that's/Users/YourName/. On Windows, it'sC:\Users\YourName\. You can open these files in any text editor — even TextEdit or Notepad.Can't find the file? If it doesn't exist yet, create it. The app will pick it up on its next restart. Make sure the entire file is valid JSON (watch for missing commas or brackets).
Claude Code users: You can skip manual editing entirely. Just run the
claude mcp addcommand above and it handles everything for you.
Step 3: Connect to Figma Desktop
Desktop Bridge Plugin:
- Open Figma Desktop normally (no special flags needed) and open a file
- Go to Plugins → Development → Import plugin from manifest...
- Select
~/.figma-console-mcp/plugin/manifest.json(stable path, auto-created by the MCP server). Runnpx figma-console-mcp --print-pathto print the exact path for your machine. - Run the plugin in your Figma file — it scans ports 9223–9232 and connects automatically to your running MCP server
Can't see the folder in the import dialog?
.figma-console-mcpstarts with a dot, so macOS hides it. In the dialog, press ⌘⇧G and paste the path (or press ⌘⇧. to show hidden files). On Windows the folder is visible as normal.
Heads-up on plugin updates. Figma caches plugin files (
code.jsandui.html) at the application level. The MCP server refreshes the files at~/.figma-console-mcp/plugin/on every startup, but Figma keeps using its cached copy until you re-import the manifest.Re-importing is required only when a release notes entry says so — typically when the plugin adds a new method the server needs (e.g. v1.22.4, v1.10.0). The plugin files last changed in v1.41.5; if your imported plugin predates that, re-import once. The plugin shows an update banner when the server bundles a newer plugin than the one running. For most upgrades the new server stays wire-compatible with the previous plugin, and re-importing is optional: you'll still get every functional change, just not the cosmetic plugin-side touches (status-pill copy,
pluginVersionreporting).When you do re-import: Plugins → Manage plugins → re-import
~/.figma-console-mcp/plugin/manifest.json. The stable path never changes, so it's a one-click step.
Step 4: Restart Your MCP Client
Restart your MCP client to load the new configuration.
Step 5: Test It!
Check Figma status→ Should show connection status with active WebSocket transport
Create a simple frame with a blue background→ Should create a frame in Figma (confirms write access!)
For Contributors: Local Git Mode
Best for: Developers who want to modify source code or contribute to the project.
What you get: Same 130 tools as NPX, plus full source code access.
Quick Setup
# Clone and build
git clone https://github.com/southleft/figma-console-mcp.git
cd figma-console-mcp
npm install
npm run build:localConfigure Your MCP Client
Add to your config file (see Where to find your config file):
{
"mcpServers": {
"figma-console": {
"command": "node",
"args": ["/absolute/path/to/figma-console-mcp/dist/local.js"],
"env": {
"FIGMA_ACCESS_TOKEN": "figd_YOUR_TOKEN_HERE",
"ENABLE_MCP_APPS": "true"
}
}
}
}Then follow NPX Steps 3-5 above.
📡 Remote SSE (Read-Only Exploration)
Best for: Quickly evaluating the tool or read-only design data extraction.
What you get: the read-only tools — view file data, components, styles, comments and version history, take screenshots, read logs, check design-code parity. Cannot create or modify designs.
Claude Desktop (UI Method)
- Open Claude Desktop → Settings → Connectors
- Click "Add Custom Connector"
- Enter:
- Name:
Figma Console (Read-Only) - URL:
https://figma-console-mcp.southleft.com/sse
- Name:
- Click "Add" — Done! ✅
OAuth authentication happens automatically when you first use design system tools.
Claude Code
⚠️ Known Issue: Claude Code's native
--transport ssehas a bug. Usemcp-remoteinstead:
claude mcp add figma-console -s user -- npx -y mcp-remote@latest https://figma-console-mcp.southleft.com/sse💡 Tip: For full capabilities, use NPX Setup instead of Remote SSE.
Other Clients (Cursor, Windsurf, etc.)
{
"mcpServers": {
"figma-console": {
"command": "npx",
"args": ["-y", "mcp-remote", "https://figma-console-mcp.southleft.com/sse"]
}
}
}Upgrading to Full Capabilities
Ready for design creation? Follow the NPX Setup guide above, or try Cloud Mode if you don't want to install Node.js.
☁️ Cloud Mode (Web AI Clients)
Best for: Using Claude.ai, v0, Replit, or Lovable to create and modify Figma designs — no Node.js required.
What you get: 105 tools including full write access — design creation, variable management, component instantiation, and all REST API tools. Only real-time monitoring (console logs, selection tracking, document changes) requires Local Mode.
Prerequisites
- [ ] Figma Personal Access Token — Create one here (starts with
figd_) - [ ] Figma Desktop with the Desktop Bridge plugin installed (see Desktop Bridge setup)
- [ ] A web AI client that supports MCP (Claude.ai, Lovable, v0, Replit, etc.)
Step 1: Add the MCP Connector
Add this endpoint to your AI platform's MCP settings:
URL: https://figma-console-mcp.southleft.com/mcp
Auth: Your Figma PAT as Bearer token
In Claude.ai: Settings → Connectors → Add Custom Connector → paste the URL above. In Lovable/v0/Replit: Look for "Add MCP Server" or "Integrations" in settings → paste the URL and add your token.
Step 2: Pair the Plugin
- Open the Desktop Bridge plugin in Figma Desktop (Plugins → Development → Figma Desktop Bridge)
- Tell your AI assistant:
Connect to my Figma plugin - The AI gives you a 6-character pairing code (expires in 5 minutes)
- In the plugin: Toggle "Cloud Mode" → enter the code → click Connect
- You're paired! Full write access is now available
What You Can Do
Once paired, use natural language to design:
Create a card component with a header image, title, description, and action button
Set up a color token collection with Light and Dark modes
Add a "High Contrast" mode to my existing token collectionHow It Works
Your AI client sends write commands through the cloud MCP server, which relays them via WebSocket to the Desktop Bridge plugin running in your Figma Desktop. The plugin executes the commands using the Figma Plugin API and returns results back through the same path.
AI Client → Cloud MCP Server → Durable Object Relay → Desktop Bridge Plugin → FigmaVariables on any plan: Cloud Mode uses the Plugin API (not the Enterprise REST API), so variable management works on Free, Pro, and Organization plans.
📊 Installation Method Comparison
| Feature | NPX (Recommended) | Cloud Mode | Local Git | Remote SSE |
|---------|-------------------|------------|-----------|------------|
| Setup time | ~10 minutes | ~5 minutes | ~15 minutes | ~2 minutes |
| Total tools | 130 | 105 after pairing | 130 (Local Git) | Read-only subset |
| Design creation | ✅ | ✅ | ✅ | ❌ |
| Variable management | ✅ | ✅ | ✅ | ❌ |
| Component instantiation | ✅ | ✅ | ✅ | ❌ |
| FigJam boards | ✅ | ✅ | ✅ | ❌ |
| Real-time monitoring | ✅ | ❌ | ✅ | ❌ |
| Desktop Bridge plugin | ✅ | ✅ | ✅ | ❌ |
| Variables (no Enterprise) | ✅ | ✅ | ✅ | ❌ |
| Console logs | ✅ (zero latency) | ❌ | ✅ (zero latency) | ✅ |
| Read design data | ✅ | ✅ | ✅ | ✅ |
| Requires Node.js | Yes | No | Yes | No |
| Authentication | PAT (manual) | OAuth (automatic) | PAT (manual) | OAuth (automatic) |
| Automatic updates | ✅ (@latest) | ✅ | Manual (git pull) | ✅ |
| Source code access | ❌ | ❌ | ✅ | ❌ |
Key insight: Remote SSE is read-only. Cloud Mode adds write access for web AI clients without Node.js. NPX/Local Git give the full 130 tools.
🎯 Test Your Connection
After setup, try these prompts:
Basic test (all modes):
Navigate to https://www.figma.com and check statusDesign system test (requires auth):
Get design variables from [your Figma file URL]Cloud Mode test:
Connect to my Figma plugin→ Follow the pairing flow, then try: "Create a simple blue rectangle"
Plugin test (Local Mode only):
Show me the primary font for [your theme name]🔐 Authentication
Remote Mode - OAuth (Automatic)
When you first use design system tools:
- Browser opens automatically to Figma authorization page
- Click "Allow" to authorize (one-time)
- Token stored securely and refreshed automatically
- Works with Free, Pro, and Enterprise Figma plans
Local Mode - Personal Access Token (Manual)
- Visit https://help.figma.com/hc/en-us/articles/8085703771159-Manage-personal-access-tokens
- Generate token with scopes (all Read unless noted): File content, File versions, Library content, Library assets, Comments (Read and write), Current user, and Variables (Enterprise only, optional)
- Add to MCP config as
FIGMA_ACCESS_TOKENenvironment variable
🛠️ Available Tools
Status & Diagnostics
figma_get_status- Check WebSocket bridge connection and file contextfigma_diagnose- Designer-readable health check + setup guidancefigma_reconnect- Force reconnect to the Desktop Bridge pluginfigma_navigate- Switch the active file target among connected plugins (Local), or navigate the cloud headless browser (Remote/Cloud)figma_list_open_files- List files connected through the Desktop Bridge and which one is active (Local Mode)
Console Debugging
figma_get_console_logs- Retrieve console logsfigma_watch_console- Real-time log streamingfigma_clear_console- Clear log bufferfigma_reload_plugin- Reload current page
Visual Debugging
figma_take_screenshot- Export an image of a page or node (Desktop Bridge first, REST fallback)figma_capture_screenshot- Render a node from the plugin runtime, reflecting changes that haven't reached the REST API yet
👁️ Real-Time Awareness (Local Mode only)
figma_get_selection- The nodes currently selected in Figmafigma_get_design_changes- Buffered document-change events, for polling what changed
Design System Extraction
figma_get_design_system_kit- Full design system in one call — tokens, components, styles, visual specsfigma_get_variables- Extract design tokens/variablesfigma_get_component- Get component data (metadata or reconstruction spec)figma_get_component_for_development- Component + imagefigma_get_component_image- Just the imagefigma_get_styles- Color, text, effect stylesfigma_get_file_data- Full file structurefigma_get_file_for_plugin- Optimized file datafigma_get_design_system_summary/figma_get_token_values- Compact design-system overview and variable values by mode (Local Mode only)figma_audit_design_system_report- Scored health audit for any MCP client — six-category report (naming, tokens, metadata, accessibility, consistency, coverage) with per-finding remediation ("can this MCP fix it?"), chunked per-category drill-down, live-first data with disclosed source, 5-minute cache
📚 Shared Library Inspection
figma_get_library_component_by_key- Resolve any component key to full properties + variants + visual specs — without needing the source library file's URL. Works for both COMPONENT_SET and standalone COMPONENT keys. Adaptive compression at >500KB.figma_get_library_components- Discover all components in a library file (requires library file URL/key)figma_get_library_variables- List every variable from team libraries the current file has subscribed. Works on every Figma plan — uses the Plugin API path, not the Enterprise-only REST endpoint. Filter bylibraryName,collectionName, orresolvedType.figma_import_library_variable- Import a library variable into the current file. Returns a localidready to pass tofigma_set_fills/figma_update_variable/ any variable-binding tool.
☁️ Cloud Relay
figma_pair_plugin- Generate a pairing code to connect a Desktop Bridge plugin via the cloud relay
✏️ Design Creation (Local Mode + Cloud Mode)
figma_execute- Power tool: Run any Figma Plugin API code to create designs- Create frames, shapes, text, components
- Apply auto-layout, styles, effects
- Build complete UI mockups programmatically
- Optional
fileKeytargets one specific connected file directly (Local Mode only), without touching the active file or target lock
figma_execute_across_files- Local Mode only. Run the same script in several Desktop Bridge-connected files at once, concurrently, returning a per-file result map — for cross-file consistency checks and fixes across a multi-file design system, replacing "open file, run plugin, repeat per file"- Name the files with
fileKeys(fromfigma_list_open_files), or passallFiles: trueto hit every connected file — one of the two is required, so a script never fans out to files you didn't mean to touch - Per-file timeouts and error isolation: one slow or failing file doesn't delay or fail the others
- Name the files with
figma_create_component_set- Create a component set with variants in one declarative call- Generate every variant combination from an axes matrix (e.g.
{ State: ["default", "hover", "disabled"], Size: ["sm", "lg"] }→ 6 variants) off a base component, or combine existing components Prop=Valuevariant naming,combineAsVariantsunder the hood, optional auto-arranged labeled grid- Returns each variant's key, ready for
figma_instantiate_component
- Generate every variant combination from an axes matrix (e.g.
figma_arrange_component_set- Organize variants into professional component sets- Convert multiple component variants into a proper Figma component set
- Applies native purple dashed border visualization automatically
- Creates white container frame with title, row labels, and column headers
- Row labels vertically centered with each grid row
- Column headers horizontally centered with each column
- Use natural language like "arrange these variants" or "organize as component set"
figma_set_description- Document components with rich descriptions- Add descriptions to components, component sets, and styles
- Supports markdown formatting for rich documentation
- Descriptions appear in Dev Mode for developers
🧩 Figma Slots (Local Mode + Cloud Mode)
figma_create_slot- Add a slot to a component via the GAcreateSlot()API — the linked SLOT property is created automatically; works on standalone components and variants inside a component setfigma_get_slots- List slots on a component, component set (aggregated across variants), or instance — ids, names, property keys, dimensions, and current childrenfigma_append_to_slot- Populate an instance's slot by cloning an existing node or creating new content (setPropertiesrejects slot values by design — this is the population path)figma_reset_slot- Clear a slot's content on an instancefigma_add_slot_property- Retrofit an existing frame as a slot via a manual SLOT property binding, withdescriptionandpreferredValues
🎬 Motion (Local Mode + Cloud Mode)
Figma Motion (Plugin API beta). Animations play in Figma's preview and in video export, not in static screenshots.
figma_get_animations- Read a layer's animation styles, keyframe tracks and timeline, with CSS easings and token bindingsfigma_apply_animation- Animate a layer with a Figma preset (Opacity, Scale, Position, Rotation, Size, Path, TextStagger), a custom animation style, or manual keyframes for any animatable property, paint color or effect. Easing and duration can bind to EASING/TIMING variablesfigma_remove_animation- Remove one animation style, one keyframe track, or everythingfigma_manage_animation_styles- Build reusable motion styles: list Figma's presets and their options, and create, update or delete custom animation stylesfigma_export_animation- Export an animated frame as MP4, GIF or WebM
🧱 Code Layers (Local Mode + Cloud Mode)
figma_get_code_layers- Find Figma code layers on a page or across the file, with position, size and linked dev resources. Figma's APIs don't expose code-layer source, so creating and editing them stays in Figma or Figma's own MCP (send_code_to_figma)
✨ Shaders (Local Mode + Cloud Mode)
figma_list_shaders- List shaders applied in the file (with each property's name, type and default, and the layers using them) and shaders from your librariesfigma_apply_shader- Apply a shader or change its properties by name ({turbulence: 40, cloudColor: '#E8F4FF'}) on fills, strokes or effects, including variable bindings. Shaders themselves are created in Figma or with Figma's own MCP
📐 Layout (Local Mode + Cloud Mode)
figma_set_layout- Auto layout and grid layout in one call: direction, gaps, padding, alignment includingSPACE_EVENLY/SPACE_AROUND, wrap, sizing, and grid tracks written CSS-style ('240px 1fr 1fr',repeat(3, 1fr)) with child placement and spans
🔍 Design-Code Parity (All Modes)
figma_check_design_parity- Compare Figma component specs against code implementation, producing a scored diff report with actionable fix itemsfigma_generate_component_doc- Generate platform-agnostic markdown documentation by merging Figma design data with code-side info; the optionalhistoryparameter adds a changelog from Figma version history andgit log
🔬 Components, Nodes & Deep Analysis (Local Mode + Cloud Mode unless noted)
figma_search_components/figma_get_component_details- Find components (local + library) and get their variants, properties, and keys (Local Mode only)figma_instantiate_component/figma_set_instance_properties- Place instances and set their TEXT, BOOLEAN, INSTANCE_SWAP, and VARIANT propertiesfigma_add_component_property/figma_edit_component_property/figma_delete_component_property- Manage component propertiesfigma_create_child,figma_move_node,figma_resize_node,figma_clone_node,figma_rename_node,figma_delete_node,figma_set_text,figma_set_fills,figma_set_strokes,figma_set_image_fill- Structured node edits, including variable binding on fills and strokes, OKLCH/Display P3 and other CSS colors on fills and strokes, and text wrap style and variable-font axes on textfigma_get_component_for_development_deep- Full component tree with resolved token names and instance referencesfigma_analyze_component_set- A variant set as a state machine, with CSS pseudo-class mappings and per-variant diffs
♿ Accessibility (Local Mode + Cloud Mode)
figma_lint_design- WCAG and design-system lint on the canvas, with AA and best-practice level taggingfigma_audit_component_accessibility- Component scorecard: state coverage, focus indicators, color-blind simulationfigma_scan_code_accessibility- axe-core scan of HTML
📝 Annotations & Comments
figma_get_annotations/figma_set_annotations/figma_get_annotation_categories- Read and write designer annotations (Local Mode + Cloud Mode)figma_get_comments/figma_post_comment/figma_delete_comment- File comments, optionally pinned to a node
🕰 Version History
figma_get_file_versions/figma_get_file_at_version- List versions and snapshot a file or node at a past versionfigma_diff_versions/figma_get_changes_since_version- Structured diffs, including component property and binding changesfigma_generate_changelog- Markdown changelog for release notesfigma_blame_node- Find when, and by whom, a property or variant was introduced
🔁 Token Sync (Local Mode + Cloud Mode)
figma_export_tokens- Export Figma variables to design token files in your codebase. Canonical DTCG JSON (legacy hex dialect by default, or DTCG 2025.10 object colors/dimensions viadtcgDialect: "2025") plus CSS, Tailwind v4/v3, SCSS, TS, JSON, Style Dictionary, and Tokens Studio formats. Diff-aware merge against existing source files (only writes what changed).tokens.config.jsonautodiscovery means zero-arg calls after first setup. Scopes and codeSyntax metadata round-trip via$extensions. Replaces Style Dictionary and Tokens Studio's export pipeline for popular styling methods.figma_import_tokens- Push code-side token edits back to Figma with a full apply phase. Diffs against current Figma state, then applies value updates, creates missing collections/variables, applies renames, writes real alias (VARIABLE_ALIAS) references, and — only understrategy: "replace"— deletes Figma-only variables. Round-trip safe — Figma variable IDs preserved in DTCG$extensions["figma-console-mcp"]so renames on either side don't create duplicates. Accepts both DTCG dialects. Dry-run strategy for safe previews. In Cloud Mode, pass tokens inline viapayloadorfiles(no local filesystem access).
🧬 Design System Extraction (Local Mode only)
Turn a production codebase into a design system: analyze → extract tokens → scaffold → port components → verify. Local Mode only — these tools read and write your local filesystem, so they are never registered in Cloud Mode.
figma_ds_analyze- Scan one or more production app codebases — framework/styling/vendor-layer detection; a usage-ranked component inventory classified vendored / wrapped / pure-vendor / bespoke, with prop contracts, variant inference from real call sites, and duplicate detection; iconography + typography capture; and architecture classification (atomic levels, specializations likeFollowButton→Button, missing generic primitives — the "UI kit vs design system" analysis)figma_ds_extract_tokens- Mine the app's styling into DTCG tokens. Declared intent first — multi-mode CSS custom properties (.dark/[data-theme]and other conventions), SCSS variables, Tailwind config values, shadcn HSL triples — plus Tailwind utility-class frequency mining valued from the app's own theme, and frequency-promoted recurring raw values. Every token carries provenance (source file:line, confidence, frequency) in$extensions; the output imports straight into Figma variables viafigma_import_tokensfigma_ds_scaffold- Generate the design-system package — package skeleton, token files via the shared formatter engine, token/typography/iconography showcase docs pages, framework peer deps, and a workflow READMEfigma_ds_setup_storybook- Wire a freshly-initialized Storybook workshop to the extraction — a preview stylesheet carrying the source app's@thememapping, custom@utilitydefinitions,@layer base, and@font-facerules; self-hosted font copying; config patches (Tailwind vite plugin, automatic JSX runtime); and a theme toolbar built from the extracted modesfigma_ds_extract_component- Deep-extract one component for porting — source, prop contract, observed call-site variants, style touchpoints, and a ready-to-adapt CSF3 story scaffoldfigma_ds_verify- Deterministic fidelity evals — DTCG parse + alias integrity, quoted-CSS-expression scan, workshopvar()resolution, structure + porting coverage, and Figma round-trip readinessfigma_ds_status- Read or record porting progress, persisted across sessions
🔧 Variable Management (Local Mode + Cloud Mode)
figma_create_variable_collection- Create new variable collections with modesfigma_create_variable- Create COLOR, FLOAT, STRING, or BOOLEAN variablesfigma_update_variable- Update variable values in specific modesfigma_rename_variable- Rename variables while preserving valuesfigma_delete_variable- Delete variablesfigma_delete_variable_collection- Delete collections and all their variablesfigma_add_mode- Add modes to collections (e.g., "Dark", "Mobile")figma_rename_mode- Rename existing modesfigma_batch_create_variables- Create up to 100 variables in one call (10-50x faster)figma_batch_update_variables- Update up to 100 variable values in one callfigma_setup_design_tokens- Create complete token system (collection + modes + variables) atomically — values accept DTCG brace references ("{color.blue.600}") that resolve to real variable aliases
📌 FigJam Board Tools (Local Mode + Cloud Mode)
figjam_create_sticky- Create a sticky note with color optionsfigjam_create_stickies- Batch create up to 200 stickiesfigjam_create_connector- Connect nodes with labeled connector linesfigjam_create_shape_with_text- Create flowchart shapes (diamond, ellipse, etc.)figjam_create_table- Create tables with cell datafigjam_create_code_block- Add code snippets with syntax highlightingfigjam_create_section- Group board content in a sectionfigjam_auto_arrange- Arrange nodes in grid, horizontal, or vertical layoutsfigjam_get_board_contents- Read all content from a FigJam boardfigjam_get_connections- Read the connection graph (flowcharts, relationships)
🎞️ Slides Presentation Tools (Local Mode + Cloud Mode)
figma_list_slides- List all slides with IDs, positions, and skip statusfigma_get_slide_content- Get the full content tree of a slidefigma_get_slide_grid- Get the 2D grid layout of the presentationfigma_get_slide_transition- Read transition settings for a slidefigma_get_focused_slide- Get the currently focused slidefigma_create_slide- Create a new blank slidefigma_delete_slide- Delete a slide from the presentationfigma_duplicate_slide- Clone an existing slidefigma_reorder_slides- Reorder slides via new 2D grid layoutfigma_set_slide_transition- Set transition effects (22 styles, 8 curves)figma_skip_slide- Toggle whether a slide is skipped in presentation modefigma_add_text_to_slide- Add text to a slide with custom fonts, colors, alignment, and wrappingfigma_add_shape_to_slide- Add rectangle or ellipse shapes with colorfigma_set_slide_background- Set a slide's background color (creates or updates)figma_get_text_styles- Get all local text styles with IDs, fonts, and sizesfigma_set_slides_view_mode- Toggle grid vs. single-slide viewfigma_focus_slide- Navigate to a specific slide
📖 Example Prompts
Cloud Mode (Web AI Clients)
Connect to my Figma plugin so we can start designing
Pair with my Figma file and create a login form with email, password, and submit button
Set up a brand color token collection with Light and Dark modesPlugin Debugging
Navigate to my Figma plugin and show me any console errors
Watch the console for 30 seconds while I test my plugin
Get the last 20 console logsDesign System Extraction
Get all design variables from https://figma.com/design/abc123
Extract color styles and show me the CSS exports
Get the Button component with a visual reference image
Get the Badge component in reconstruction format for programmatic creationDesign Creation (Local Mode + Cloud Mode)
Create a success notification card with a checkmark icon and message
Design a button component with hover and disabled states
Build a navigation bar with logo, menu items, and user avatar
Create a modal dialog with header, content area, and action buttons
Arrange these button variants into a component set
Organize my icon variants as a proper component set with the purple borderVariable Management (Local Mode + Cloud Mode)
Create a new color collection called "Brand Colors" with Light and Dark modes
Add a primary color variable with value #3B82F6 for Light and #60A5FA for Dark
Rename the "Default" mode to "Light Theme"
Add a "High Contrast" mode to the existing collectionDesign-Code Parity
Compare the Button component in Figma against our React implementation
Check design parity for the Card component before sign-off
Generate component documentation for the Dialog from our design systemFigJam Boards
Create a retrospective board with "Went Well", "To Improve", and "Action Items" columns
Build a user flow diagram for the checkout process with decision points
Read this brainstorming board and summarize the key themes
Generate an affinity map from these meeting notes
Create a comparison table of our three platform optionsSlides Presentations
List all slides and tell me which ones are skipped
Add a new slide with the title "Thank You" in 72px text
Set a DISSOLVE transition on the first slide with 0.5 second duration
Duplicate slide 5 for an A/B comparison
Skip slides 8 and 9 — they're not ready for the client presentation
Reorder my slides so the conclusion comes before Q&AVisual Debugging
Take a screenshot of the current Figma canvas
Navigate to this file and capture what's on screen🎨 AI-Assisted Design Creation
Requires Desktop Bridge: This feature works with Local Mode (NPX or Local Git) and Cloud Mode. Remote SSE without Cloud Mode pairing is read-only and cannot create or modify designs.
One of the most powerful capabilities of this MCP server is the ability to design complete UI components and pages directly in Figma through natural language conversation with any MCP-compatible AI assistant like Claude Desktop or Claude Code.
What's Possible
Create original designs from scratch:
Design a login card with email and password fields, a "Forgot password?" link,
and a primary Sign In button. Use 32px padding, 16px border radius, and subtle shadow.Leverage existing component libraries:
Build a dashboard header using the Avatar component for the user profile,
Button components for actions, and Badge components for notifications.Generate complete page layouts:
Create a settings page with a sidebar navigation, a main content area with form fields,
and a sticky footer with Save and Cancel buttons.How It Works
- You describe what you want in plain English
- The AI searches your component library using
figma_search_componentsto find relevant building blocks - Components are instantiated with proper variants and properties via
figma_instantiate_component - Custom elements are created using the full Figma Plugin API via
figma_execute - Visual validation automatically captures screenshots and iterates until the design looks right
Who Benefits
| Role | Use Case | |------|----------| | Designers | Rapidly prototype ideas without manual frame-by-frame construction. Explore variations quickly by describing changes. | | Developers | Generate UI mockups during planning discussions. Create visual specs without switching to design tools. | | Product Managers | Sketch out feature concepts during ideation. Communicate visual requirements directly to stakeholders. | | Design System Teams | Test component flexibility by generating compositions. Identify gaps in component coverage. | | Agencies | Speed up initial concept delivery. Iterate on client feedback in real-time during calls. |
Example Workflows
Brand New Design:
"Create a notification toast with an icon on the left, title and description text, and a dismiss button. Use our brand colors."
The AI creates custom frames, applies your design tokens, and builds the component from scratch.
Component Composition:
"Build a user profile card using the Avatar component (large size), two Button components (Edit Profile and Settings), and a Badge for the user's status."
The AI searches your library, finds the exact components, and assembles them with proper spacing and alignment.
Design Iteration:
"The spacing feels too tight. Increase the gap between sections to 24px and make the heading larger."
The AI modifies the existing design, takes a screenshot to verify, and continues iterating until you're satisfied.
Visual Validation
The AI automatically follows a validation workflow after creating designs:
- Create → Execute the design code
- Screenshot → Capture the result
- Analyze → Check alignment, spacing, and visual balance
- Iterate → Fix any issues detected
- Verify → Final screenshot to confirm
This ensures designs aren't just technically correct—they look right.
🎨 Desktop Bridge Plugin (Recommended Connection)
The Figma Desktop Bridge plugin is the recommended way to connect Figma to the MCP server. It communicates via WebSocket — no special Figma launch flags needed, and it persists across Figma restarts.
Setup
- Open Figma Desktop (normal launch — no debug flags needed)
- Go to Plugins → Development → Import plugin from manifest...
- Select
~/.figma-console-mcp/plugin/manifest.json— the stable path the MCP server keeps up to date (npx figma-console-mcp --print-pathprints it; on macOS press ⌘⇧G in the dialog and paste it, since the folder is hidden). If you cloned this repo for development, you can selectfigma-desktop-bridge/manifest.jsonfrom the clone instead. - Run the plugin in your Figma file — it auto-connects via WebSocket (scans ports 9223–9232)
- Ask your AI: "Check Figma status" to verify the connection
One-time import. Once imported, the plugin stays in your Development plugins list. Just run it whenever you want to use the MCP.
📖 Desktop Bridge Documentation
Capabilities
Read Operations:
- Variables without Enterprise API
- Reliable component descriptions (bypasses API bugs)
- Multi-mode support (Light/Dark/Brand variants)
- Real-time selection tracking and document change monitoring
Write Operations:
- Design Creation - Create frames, shapes, text, components via
figma_execute - Variable Management - Full CRUD operations on variables and collections
- Mode Management - Add and rename modes for multi-theme support
How the Transport Works
- The MCP server communicates via WebSocket through the Desktop Bridge plugin
- The server tries port 9223 first, then automatically falls back through ports 9224–9232 if needed
- The plugin scans all ports in the range and connects to every active server it finds
- All 130 tools work through the WebSocket transport
Multiple files: The WebSocket server supports multiple simultaneous plugin connections — one per open Figma file. Each connection is tracked by file key with independent state (selection, document changes, console logs).
Environment variables:
FIGMA_WS_PORT— Override the preferred WebSocket port (default: 9223). The server will fall back through a 10-port range starting from this value if the preferred port is occupied.FIGMA_WS_HOST— Override the WebSocket server bind address (default:localhost). Set to0.0.0.0when running inside Docker so the host machine can reach the MCP server.FIGMA_MCP_HTTP_PORT— Also serve MCP over Streamable HTTP athttp://127.0.0.1:<port>/mcp, so other local clients can share this server's Figma connection. Off by default. See Sharing one server over HTTP.FIGMA_MCP_HTTP_TOKEN— WithFIGMA_MCP_HTTP_PORT, requireAuthorization: Bearer <token>on every HTTP request.
Cloud Mode: The plugin also supports a Cloud Mode toggle for pairing with web AI clients (Claude.ai, v0, Replit, Lovable). Toggle "Cloud Mode" in the plugin UI, enter the 6-character pairing code from your AI assistant, and click Connect. See Cloud Mode for details.
Plugin Limitation: In Local Mode, works with NPX or Local Git. In Cloud Mode, pairs with the remote MCP endpoint. Remote SSE without Cloud Mode pairing is read-only.
🔀 Multi-Instance Support (v1.10.0)
Figma Console MCP now supports multiple simultaneous instances — perfect for designers and developers who work across multiple projects or use Claude Desktop's Chat and Code tabs at the same time.
The Problem (Before v1.10.0)
When two processes tried to start the MCP server (e.g., Claude Desktop's Chat tab and Code tab), the second one would crash with EADDRINUSE because both competed for port 9223.
How It Works Now
- The server tries port 9223 first (the default)
- If that port is already taken, it automatically tries 9224, then 9225, and so on up to 9232
- The Desktop Bridge plugin in Figma connects to all active servers simultaneously
- Every server instance receives real-time events (selection changes, document changes, console logs)
figma_get_statusshows which port you're on and lists other active instances
What This Means for You
| Scenario | Before v1.10.0 | Now | |----------|----------------|-----| | Two Claude Desktop tabs (Chat + Code) | Second tab crashes | Both work independently | | Multiple CLI terminals on different projects | Only one can run | All run simultaneously | | Claude Desktop + Claude Code CLI | Port conflict | Both coexist |
Do I Need to Do Anything?
Nothing. Multi-instance support is fully automatic:
- Each MCP server claims the next available port in the range
- The Desktop Bridge plugin scans all ports and connects to every active server
- Orphaned processes from closed tabs are automatically cleaned up on startup
- No manual port management — the plugin already scans the whole range
(Re-importing the manifest is only required when the plugin code itself changes — e.g. after a package update. Port-range scanning is already in the shipped plugin.)
Sharing one server over HTTP (optional)
Normally each MCP client starts its own server over stdio. To let several local clients share one server and its Figma connection, set FIGMA_MCP_HTTP_PORT. The server then also serves MCP at http://127.0.0.1:<port>/mcp, and lists that URL in its port file (mcpHttpUrl) and in figma_get_status (transport.mcpHttp).
# Start a shared server in a terminal (keep it open)
FIGMA_ACCESS_TOKEN=figd_... FIGMA_MCP_HTTP_PORT=9300 npx -y figma-console-mcp@latest
# Point clients at it, e.g. Claude Code:
claude mcp add --transport http figma-console-shared http://127.0.0.1:9300/mcp- Loopback only. It binds
127.0.0.1and refuses requests whoseHostisn't localhost (DNS rebinding) or that come from a non-localhost browser page. SetFIGMA_MCP_HTTP_TOKENto also require a bearer token. - Each client gets its own session, as with stdio, and all of them share the plugin connection. Idle sessions close after 30 minutes and clients reconnect automatically.
- The endpoint lives as long as the server process. When an MCP client spawned the server, closing that client ends it.
- Avoid port 3845, which Figma's own Dev Mode MCP server uses, and the bridge range 9223–9232.
🧩 MCP Apps (Experimental)
Figma Console MCP includes support for MCP Apps — rich interactive UI experiences that render directly inside any MCP client that supports the MCP Apps protocol extension. Built with the official @modelcontextprotocol/ext-apps SDK.
What are MCP Apps? Traditional MCP tools return text or images to the AI. MCP Apps go further — they render interactive HTML interfaces inline in the chat, allowing users to browse, filter, and interact with data directly without consuming AI context.
Token Browser
An interactive design token explorer.
Usage: Ask Claude to "browse the design tokens" or "show me the design tokens" while connected to a Figma file.
Features:
- Browse all tokens organized by collection with expandable sections
- Filter by type (Colors, Numbers, Strings) and search by name/description
- Per-collection mode columns (Light, Dark, Custom) matching Figma's Variables panel
- Color swatches, alias resolution, and click-to-copy on any value
- Works without Enterprise plan via Desktop Bridge (local mode)
Design System Dashboard
A Lighthouse-style health scorecard that audits your design system across six categories.
Usage: Ask Claude to "audit the design system" or "show me design system health" while connected to a Figma file.
Features:
- Overall weighted score (0–100) with six category gauges: Naming, Tokens, Components, Accessibility, Consistency, Coverage
- Expandable category sections with individual findings, severity indicators, and actionable details
- Diagnostic locations linking findings to specific variables, components, or collections
- Tooltips explaining each check's purpose and scoring criteria
- Refresh button to re-run the audit without consuming AI context
- Pure scoring engine with no external dependencies — all analysis runs locally
No MCP Apps support? Same audit, plain tool: figma_audit_design_system_report runs the identical scoring engine and returns the report as data — summary by default, per-category drill-down via category, full JSON via format: "full" — with a remediation section stating which findings this MCP can fix (design), which need a design decision first (design-assisted), and which need human design work (manual). Works in every MCP client with no ENABLE_MCP_APPS flag; results cache for 5 minutes (forceRefresh to re-crawl).
Enabling MCP Apps:
MCP Apps are enabled by default in the setup configurations above (via "ENABLE_MCP_APPS": "true"). If you set up before v1.10.0 and don't have this in your config, add it to your env section:
"env": {
"FIGMA_ACCESS_TOKEN": "figd_YOUR_TOKEN_HERE",
"ENABLE_MCP_APPS": "true"
}Note: MCP Apps require an MCP client with ext-apps protocol support (e.g. Claude Desktop). This feature is experimental and the protocol may evolve.
Future MCP Apps Roadmap
Planned MCP Apps:
- Component Gallery — Visual browser for searching and previewing components with variant exploration
- Style Inspector — Interactive panel for exploring color, text, and effect styles with live previews
- Variable Diff Viewer — Side-by-side comparison of token values across modes and branches
The architecture supports adding new apps with minimal boilerplate — each app is a self-contained module with its own server-side tool registration and client-side UI.
🚀 Advanced Topics
- Setup Guide - Complete setup guide for all MCP clients
- Self-Hosting - Deploy your own instance on Cloudflare
- Architecture - How it works under the hood
- OAuth Setup - Configure OAuth for self-hosted deployments
- Troubleshooting - Common issues and solutions
🤝 vs. Figma's Official MCP Server
The two servers are complementary, and both can run in the same MCP client.
Figma's official MCP server is hosted by Figma. It focuses on design-to-code context (get_design_context), Code Connect, library search, and creating content in Figma: it writes through use_figma, creates files, generates FigJam diagrams, generates code layers, and works with shaders.
Figma Console MCP focuses on design-system operations:
- Deterministic checks: design-code parity, accessibility (canvas and code), and design-system health
- Token export in 10 formats and DTCG import with round-trip variable IDs
- Variable reads and writes on any Figma plan through the Desktop Bridge plugin
- Component documentation with design and code history, and version-history diffs
- Codebase → design system extraction
- Plugin console capture, selection tracking, and multi-file execution (Local Mode)
🛤️ Roadmap
Current Status: v1.46.0 (Stable) - Production-ready. Latest: shaders applied with properties set by name, figma_set_layout for auto layout and grid layout, OKLCH/Display P3 and other CSS colors on fills and strokes, and text wrap style and variable-font axes in figma_set_text (server-only). On top of v1.45.1's figma_get_code_layers no longer hangs on a file in a background window, and the tool docs were audited against every tool's real parameters (server-only). On top of v1.45.0's Motion tools — apply, read, remove and export Figma Motion animations, manage custom animation styles, write Timing and Easing variables from token files, and find code layers (server-only). On top of v1.44.2's Cloud sign-in requests library_assets:read, so component lookups by key work over OAuth (server-only). On top of v1.44.1's Cloud Mode accepting personal access tokens without the Current user scope, 403 errors name the missing scope, and setup guides for Codex, Gemini CLI and Antigravity (server-only). On top of v1.44.0's opt-in Streamable HTTP endpoint for Local Mode (FIGMA_MCP_HTTP_PORT, loopback only, optional bearer token) so several local clients share one server and its Figma connection (server-only). On top of v1.43.0's @modelcontextprotocol/sdk 1.32.1 and zod 4 clear all seven production security advisories, with tool behavior unchanged and Gemini-safe schemas (server-only). On top of v1.42.1's commands to a plugin stuck on a timed-out script fail fast with a busy message instead of queuing behind it, port-exhaustion eviction ends the least recently used server instead of the oldest, and figma_execute accepts up to a 300 s timeout (server-only). On top of v1.42.0's figma_get_comments returning compact threads (root + replies, newest activity first) with node_id, thread_id and since filters and limit/offset paging, 30 threads by default; summary counts are per thread (server-only, PR #139). On top of v1.41.5's component-set analysis reporting gradient/image/paint-opacity, corner radius, typography and label text, swapped spacing tokens, nested layer geometry and vector shape changes (re-import the plugin). On top of v1.41.4's variant background/stroke and auto-layout changes. On top of v1.41.3's component-set analysis reporting changes inside nested layers (instance swaps, nested token changes, opacity on the variant itself) and no longer reports a layer hidden in every variant as a change (re-import the plugin). On top of v1.41.2's design-system summary and figma_get_token_values report the file's tokens (they always reported 0) (server-only). On top of v1.41.1's component-set analysis reports real variant differences (Figma's default variant, size and layer changes), component loading no longer times out on large files (throttled plugin timers), relevance-ranked component search, and figma_get_variables paging that works with format: "full" (re-import the plugin for the component fixes). On top of v1.41.0's Cloud Mode sign-in hardening (per-session tokens on /sse, single-use authorization codes bound to client, redirect URI and PKCE, a consent page for unrecognized destinations), a script-execution fix in the accessibility scanner, figma_import_tokens refusing empty imports and scoping replace deletes to imported collections, figma_arrange_component_set no longer deleting other arranged sets, live variable reads for token export/import, Windows orphan-server exit (PR #140), and fileKey for the Slots tools (PR #112) (server-only, no plugin re-import). On top of v1.40.9's composed colors (a color variable set to another color at an opacity, "grey/900 at 50%") export and import with the link to the primitive intact — color-mix() in CSS/Tailwind v4, rgba() in SCSS, an alpha modifier in Tokens Studio, the resolved color plus the composition in DTCG — where the export previously failed or wrote [object Object] (server-only, no plugin re-import). On top of v1.40.8's min/max width and height survive every way of extracting a component as JSON — the design system kit, the reconstruction format and figma_get_file_data now carry them, design-code parity compares them, and the reconstruction format returns real geometry again instead of 50×50 placeholders when the plugin is connected (server-only, no plugin re-import). On top of v1.40.7's clearer Color Tokens table in generated docs — layers revealed by a boolean property are printed once under that property, and stroke/fill rows name the nested instance they belong to (server-only, no plugin re-import). On top of v1.40.6's overrides in extended variable collections are visible (previously reported as nonexistent), and a third round of generated-doc fidelity fixes — radius tokens read per corner, boolean props compared in Design-Code Parity, source links pinned to a commit, hidden layers linked to the property that shows them (server-only, no plugin re-import). On top of v1.40.5's clean dependency audit for the npm package — Worker-only dependencies (@cloudflare/puppeteer, which pulled in a high-severity extract-zip advisory with no upstream fix, and agents) no longer install for npm users, so npm install figma-console-mcp audits at 0 vulnerabilities and is 64 MB instead of 173 MB; the server also reports its real version to MCP clients instead of 0.1.0 (server-only, no plugin re-import). On top of v1.40.4's fixes found by live-testing v1.40.3 on hard production components — tools given a fileUrl now read variables, descriptions and annotations from THAT file rather than whichever file is active (previously every token name could vanish, or on an id collision come from the wrong file; figma_get_component could return a different component), a tool result over 16 MB no longer disconnects the server, anatomy labels sizing by the layout's real axes, and the icon/typography/anatomy output stays readable on large components (server-only, no plugin re-import). On top of v1.40.3's fidelity fixes for generated component docs on harder components — hidden layers are labeled instead of documented as rendered, one-sided borders report the weight that actually renders, typography and layer structure are compared across every variant instead of read from the first, gradients/shadows/opacity are no longer silently dropped, truncated trees say so, and SLOT properties are listed (server-only, no plugin re-import). On top of v1.40.2's data-loss fix for figma_export_tokens — with a different file active in Figma than intended, an export could replace a real token file with an empty one (or with another file's tokens) and report success; exports now refuse to write when a requested collection is missing, when the export is empty, when the target was generated from a different Figma file, or when overwriting would destroy tokens the export doesn't manage, and strategy: "merge" — documented as preserving code-only tokens but never implemented — is now a genuinely safe default (server-only, no plugin re-import). On top of v1.40.1's accuracy fixes for generated component docs across variants — a descendant's fill (typically an icon's) could be documented as a variant's Background, Color Token headings kept only a property literally named Variant, spacing was read from the first variant alone, and icons were only detected by layer name — plus figma_export_tokens accepting a file path as outputPath instead of failing with a raw EEXIST or silently creating a directory (server-only, no plugin re-import). On top of v1.40.0's design system extraction from production codebases — seven Local-Mode figma_ds_* tools that scan one or more apps (framework/styling/vendor detection, usage-ranked component inventory classified vendored/wrapped/pure-vendor/bespoke, variant inference from real call sites, duplicate detection, and an architecture pass separating UI-kit specializations from the missing generic primitives), mine the de-facto styling into DTCG tokens with per-token provenance (multi-mode CSS custom properties, SCSS, Tailwind config, shadcn HSL triples, utility-class frequency mining valued from the app's own theme), scaffold the design-system package with showcase docs pages, wire a fresh Storybook workshop to the app's real theme layers and fonts, deep-extract components for porting with CSF3 story scaffolds, verify with deterministic fidelity evals, and persist porting progress across sessions — the extracted tokens import straight into Figma variables via figma_import_tokens. On top of v1.39.1's fix for a plugin-update banner that could never be cleared — an older server instance nagging a newer plugin to re-import, which only ever installs the same or a newer plugin. On top of v1.39.0's multi-file execution — figma_execute_across_files runs one script against several Desktop Bridge-connected files concurrently (four files x 2s of work in ~3s, not ~8s), and figma_execute takes an optional fileKey to target a single file without moving the active file or releasing target lock; targeting must be explicit (fileKeys or allFiles: true), so nothing fans out to files you did not name. From community PR #107 by @Wolfr. Ships with a relay fix: handleResult() in the plugin's ui.html rebuilt each response field by field and silently dropped resultAnalysis and fileContext, so the resultAnalysis.warning check that figma_execute's own description tells callers to perform has never been possible (re-import manifest.json for those two fields; everything else works without it). On top of v1.38.2's connection-stability fix — the server's own orphan reaper was terminating healthy MCP servers because its liveness probe used an IPv4 literal while the server binds the IPv6 loopback, which turned every kill-safety gate into a rubber stamp; this was the cause of recurring "Server disconnected" errors across all MCP clients. Alongside v1.38.1's fix for the shared-library variable tools, which had silently reported zero collections since v1.29.0 and reported failed imports as successes. On top of v1.38.0's ongoing component changelogs in generated docs — figma_generate_component_doc takes an opt-in history parameter that pulls per-component design history from Figma version history (each version diffed and scoped to that component, so renames, added properties, and token bindings show up as rows) alongside code history from git log on the component's source files, rendered as a ## History section with design, code, and release-note tables. Prefers labeled versions, falls back to auto-saves when a file has none, and is off by default so existing calls are unchanged (server-only, no plugin re-import). On top of v1.37.x's design-system health aud
