Tuval Studio MCP
This MCP server lets AI agents create, edit, audit, and export graphic designs, multi-page PDFs/catalogs, and social/presentation assets through a local browser canvas.
Check Tuval Studio connection status and capture live canvas screenshots for vision-based inspection.
Create or resize canvases, set solid backgrounds, transparency, gradients, patterns, and load templates.
Add and style text with Google Fonts, sizes, weights, alignment, color, spacing, and line height.
Add geometric shapes and SVG stickers/badges, then update fills, strokes, radius, opacity, shadows, and gradients.
Insert images from local files, URLs, or Base64; crop, auto-remove backgrounds, magic-wand remove backgrounds, and apply color filters.
Manage layers: list, reorder, align, transform, flip, rotate, set opacity/blend mode, hide/lock/rename, duplicate, delete, and add shadows.
Audit layouts for collisions, overlaps, and boundary overflows.
Manage multi-page documents: list, add, switch, duplicate, delete, and export all pages as a multi-page PDF catalog.
Search/list local assets, built-in stickers, templates, patterns, stock PNGs, and copyright-free stock elements.
Execute batch canvas actions, draw freehand brush strokes, undo/redo, rename projects, save locally, download exports, or export image files at multiple scales.
Allows adding and styling text using Google Fonts, providing access to font families such as Sora, Inter, Playfair Display, Bebas Neue, Cinzel, and Syne for graphic design projects.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@Tuval Studio MCP make a 3-page product catalog PDF with a modern gradient cover"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Tuval Studio MCP Server ๐จ
The Graphic Design MCP, AI Design MCP & PDF Generator MCP for AI Agents
Tuval Studio MCP is the official MCP server for Tuval Studio (tuval.site).
It empowers desktop AI agents โ including CodeDL, Claude Desktop, Cursor, Codex, Antigravity, Windsurf, Cline, Roo-Code, and OpenCode โ with an autonomous, vision-guided design engine for:
๐จ Graphic Design MCP & AI Design MCP: Automated vector composition, typography, mesh gradients, and smart layout audits.
๐ PDF Generator MCP & Catalog Generator: Multi-page PDF catalog generation, magazine layouts, brochures, and pitch decks.
๐ Presentation Generator & Slide Decks: High-resolution multi-page slide creation with synchronized typography and consistent visual branding.
๐ฑ Social Media Design: 1-click marketing banners, Instagram stories, YouTube thumbnails, and ads with zero text overflow.
Looking for a design MCP, PDF MCP, catalog MCP, or graphic design MCP? Tuval Studio MCP connects your AI agent directly to an interactive, client-side browser canvas with 100% privacy and real-time visual feedback!
๐ Core Features
โก Powered by APIDL Engine (apidlai.com): Breakthrough closed-loop design orchestration developed by apidlai.com. Converges in 2โ3 iterative phases (
draft->refine->final) with automated Quality Gate audits, slashing token costs by ~92% and accelerating execution by 10x compared to traditional single-call MCP tools.๐ Zero-Cloud & 100% Private: Operates entirely in your local browser through a high-speed local WebSocket bridge (
ws://127.0.0.1:8765). Zero account or server storage required.๐ Multi-Page PDF Catalog Engine: Autonomous creation, management, and export of multi-page magazines, lookbooks, product brochures, and pitch decks with
pdf_export_catalog.๐๏ธ Multimodal Vision-in-the-Loop: Every action returns high-res visual snapshots (Base64 PNG) and layer coordinates so AI agents inspect, critique, and perfect their designs visually.
๐ก๏ธ Zero-Overflow & Boundary Guard: Built-in validation protocol ensuring all typography, buttons, and badges maintain safe margins (min 50px) with no clipping.
โจ Full Graphic Suite: Multi-layer vector manipulation, custom Google Fonts (Sora, Inter, Playfair, Cinzel, Syne, etc.), mesh gradients, geometric pattern textures, and 1-click AI background removal.
๐ป Local File Loading: Load and composite any local photo, asset, or logo from your computer hard drive directly onto the canvas.
Related MCP server: pulp
โก Quick Setup for AI Agents
Add the configuration below to your AI agent's MCP settings file:
1. Claude Desktop (claude_desktop_config.json)
{
"mcpServers": {
"tuval-studio": {
"command": "npx",
"args": ["-y", "@ariferol01/tuval-studio-mcp@latest"]
}
}
}2. CodeDL, Cursor, Codex, Antigravity, Windsurf, Cline & Roo-Code
{
"mcpServers": {
"tuval-studio": {
"command": "npx",
"args": ["-y", "@ariferol01/tuval-studio-mcp@latest"]
}
}
}How it works: Open Tuval Studio in your browser tab. When your AI agent executes design commands, it connects instantly to your active canvas tab!
๐ ๏ธ Complete MCP Tool Catalog
1. Document & Multi-Page PDF Engine
Tool | Description |
| Lists all pages in the document with indices, dimensions, and active status. |
| Creates a new page (blank or cloned template) for catalog & brochure workflows. |
| Switches the active canvas viewport to a specific page index ( |
| Clones the current page layout and all layers to maintain visual consistency. |
| Deletes a page from the document. |
| Moves / reorders a page from one position to another in the multi-page filmstrip. |
| Compiles all pages into a print-ready, high-resolution multi-page PDF document. |
2. Canvas & Artboard Setup
Tool | Description |
| Checks connection status and instructs agent to open |
| Initializes a blank canvas with custom width, height, and background color. |
| Applies standard dimensions or social presets ( |
| Sets solid background color, alpha transparency, or full-bleed background textures. |
| Applies mesh gradients ( |
| Applies geometric textures ( |
| Loads curated design templates ( |
| Clears all objects from the current canvas. |
3. Typography & Vector Shapes
Tool | Description |
| Adds text with Google Fonts ( |
| Updates font family, font size, color, letter spacing, line height, text align, and content. |
| Inserts geometric shapes ( |
| Modifies shape fill color, border stroke, corner radius ( |
| Places curated badges, trust stickers, flares, stars, and callouts onto the canvas. |
4. Images, Media & Local Assets
Tool | Description |
| Reads an image from the user's local hard drive and converts it for instant canvas placement. |
| Scans and lists local images and logos available in the workspace. |
| Inserts an image from URL, local file path, or Base64 (supports |
| Crops images by bounding box coordinates or aspect ratio. |
| 1-click client-side background remover. |
| Magic wand background extraction with configurable color tolerance. |
| Applies aesthetic color grading ( |
5. Layer Management, Layout Audit & Vision Inspection
Tool | Description |
| Automated zero-overflow and collision audit across all text and image layers. |
| Captures high-res canvas snapshot (Base64 PNG) & layer tree for visual AI audit. |
| Moves layers in stack ( |
| Aligns layers to canvas center, left, right, top, or bottom. |
| Adjusts rotation angle, flip X/Y, opacity, drop shadow, and blend modes. |
| Duplicates or removes specific layers. |
| Executes sequential multi-step commands in a single atomic call. |
| Iterative Closed-Loop APIDL Pipeline: Executes batch design scripts, returns multimodal visual state + layer collision audit, enforces Golden Ratio standards, and converges in 2-3 iterations ( |
| Renames the design project. |
| Triggers browser download for PNG, JPG, WebP, or SVG graphics. |
๐ฏ Example Prompts for AI Agents (CodeDL, Claude, Cursor, Codex, Antigravity)
1. Multi-Page Product Catalog & PDF Generation
"Use Tuval Studio MCP to create a 4-page modern Scandinavian furniture catalog. Design an editorial cover on Page 1, a product grid with pricing on Pages 2 and 3, and contact/order details on Page 4. Inspect each page with vision audit for safe margins and export the final multi-page PDF."
2. Social Media Design & Banner Generation
"Design a high-contrast Cyberpunk social media ad banner (1080x1080) for a product launch. Apply dark gradient backgrounds, bold typography, neon callout badges, and verify with design_audit_layout that no text overflows boundaries."
3. Presentation Generator & Pitch Decks
"Build a 5-slide pitch deck presentation in Tuval Studio. Maintain brand colors, modern typography hierarchy, clean card layouts, and export the entire deck as a print-ready PDF."
๐ Tags & Keywords
Tuval Studio MCP ยท MCP server ยท graphic design MCP ยท AI design MCP ยท PDF generator MCP ยท PDF MCP ยท design MCP ยท catalog MCP ยท catalog generator ยท presentation generator ยท social media design ยท AI agents ยท CodeDL ยท Claude ยท Cursor ยท Codex ยท Antigravity
๐ License
MIT ยฉ Tuval Studio
Available Tools
53 toolsassets_list_localA
Scans and lists all local images, icons, logos, and graphic assets in the workspace-assets directory ready to be placed onto the canvas.
Inputs: directory.
assets_list_patternsA
Lists all built-in geometric pattern and texture background presets (Dot Matrix, Technical Blueprint, Carbon Fiber, Diagonal Stripes, 3D Isometric Wireframe, Topography, Circuit Board).
Inputs: no named parameters.
assets_list_stickersA
Lists all available built-in SVG stickers, badges, and vector elements available for insertion.
Inputs: no named parameters.
assets_list_stock_pngsB
Lists free transparent 3D cutout assets, holographic spheres, lighting flares, and curated royalty-free photography available for instant placement.
Inputs: no named parameters.
assets_load_from_pathA
Reads any local image or graphic file from the user computer (absolute path e.g. "C:/Users/name/Desktop/logo.png" or relative path), encodes it to Base64, and returns the data URL for use in canvas.
Inputs: filePath (required).
brush_drawC
Executes freehand brush drawing on the canvas with specific brush style ("pencil", "spray", "circle", "eraser"), color, size, and path coordinates.
Inputs: size, color, points, opacity, +1 more.
canvas_clearB
Clears all layers and objects from the canvas.
Inputs: no named parameters.
canvas_create_newB
Creates a new blank canvas with specific dimensions and background color.
Inputs: width, height, isTransparent, backgroundColor.
canvas_load_templateC
Loads a pre-designed layout template onto the canvas (e.g. "social-launch", "neon-sale-story", "youtube-thumbnail", "minimal-post", "bold-thumbnail").
Inputs: templateId (required).
canvas_set_backgroundA
Sets the solid background color or enables alpha transparency. NOTE FOR AI DESIGNERS: You are NOT restricted to solid colors! You can search & apply rich background photography, architectural textures, gradients, or pattern grids using "image_add" with { isBackground: true }, or choose from "assets_list_patterns" / "assets_list_stock_pngs" / "assets_list_local".
Inputs: color (required), isTransparent.
canvas_set_dimensionsC
Changes the dimensions or applies a social media preset (1080x1080 Instagram Post, 1080x1920 Story, 1280x720 YouTube Thumbnail, 1200x630 Facebook/Web Banner).
Inputs: width, height, preset.
canvas_set_gradientB
Applies a modern gradient to the canvas background. You can specify a built-in preset ("cyber", "sunset", "emerald", "space", "gold", "candy", "noir") or custom colors array and angle.
Inputs: angle, colors, preset.
canvas_set_patternC
Applies a patterned or textured geometric background to the canvas artboard.
Inputs: patternKey (required), bgColor.
design_audit_layoutA
Audits the current active canvas for text collisions, layer overlaps, and boundary overflows. Returns actionable recommendations for fixing overlaps and spacing.
Inputs: no named parameters.
element_add_stockB
Adds a free copyright-free stock image or cutout to the canvas directly by search query, image URL, or asset key.
Inputs: url (required), name, scaleToWidth.
elements_searchC
Searches through thousands of copyright-free stock photos, transparent PNG cutouts, SVG vector icons, trust badges, and backgrounds by keywords and category filters.
Inputs: page, query, category.
filter_adjust_valuesB
Manually adjusts color grading sliders on an image layer (brightness, contrast, saturation, hue, blur, noise, sepia, grayscale).
Inputs: layerId (required), hue, blur, noise, +5 more.
filter_apply_presetB
Applies an aesthetic color grading preset to an image layer ("Warm Film", "Cyberpunk", "Moody Noir", "Vintage 70s", "Pastel Soft", "Vivid HDR", "Film Grain", "Normal").
Inputs: layerId (required), presetName (required).
history_redoB
Redoes the previously undone change.
Inputs: no named parameters.
history_undoB
Reverts the canvas to the previous undo history state.
Inputs: no named parameters.
image_addA
Adds an image layer to the canvas. Source can be ANY local file path on the user computer (e.g. "C:/path/to/img.png" or "workspace-assets/photo.jpg"), an external public URL (e.g. Unsplash), or a Base64 data URL. Set isBackground: true to automatically make it a full-bleed background.
Inputs: src (required), x, y, angle, +3 more.
image_cropC
Crops an image layer with a given aspect ratio (1:1, 4:5, 16:9, 9:16) or custom pixel coordinates.
Inputs: layerId (required), cropH, cropW, cropX, +2 more.
image_remove_bg_autoC
Applies AI 1-click automatic background removal to an image layer by analyzing edges.
Inputs: layerId (required).
image_remove_bg_magicA
Performs Magic Wand background removal on an image layer by sampling color at click coordinates with tolerance and feather smoothing.
Inputs: layerId (required), x, y, feather, +1 more.
layer_alignC
Aligns an object to the canvas artboard ("center", "left", "right", "top", "bottom", "center_h", "center_v").
Inputs: layerId (required), alignment (required).
layer_deleteB
Deletes a layer from the canvas.
Inputs: layerId (required).
layer_duplicateC
Duplicates an existing layer and adds the copy to the canvas.
Inputs: layerId (required).
layer_reorderC
Reorders a layer in the visual z-index stack ("bring_to_front", "send_to_back", "bring_forward", "send_backward").
Inputs: action (required), layerId (required).
layer_set_gradientC
Applies a gradient fill to an existing text or shape layer by layerId.
Inputs: layerId (required), angle, colors, preset.
layer_set_shadowC
Adds or updates drop shadow on a layer (color, blur radius, horizontal/vertical offsets, or quick preset: "soft", "float", "glow", "none").
Inputs: layerId (required), blur, color, preset, +2 more.
layer_set_stateB
Toggles visibility (hide/show), locks/unlocks selection, or renames a layer.
Inputs: layerId (required), name, locked, visible.
layers_get_allA
Returns the complete hierarchical list of all objects/layers currently on the canvas with their IDs, positions, and styles.
Inputs: no named parameters.
layer_transformC
Transforms a layer: move X/Y, rotate angle, flip horizontally/vertically, set opacity, or change blend mode.
Inputs: layerId (required), x, y, angle, +4 more.
pages_addC
Adds a new page to the document. Ideal for designing multi-page catalogs, product lookbooks, brochures, magazines, and pitch decks.
Inputs: title, width, height, bgColor, +2 more.
pages_deleteA
Deletes a page from the document by index (the document must always keep at least 1 page).
Inputs: index (required).
pages_duplicateA
Duplicates an existing page (including all text, shapes, images, and styling) into a new page immediately following it.
Inputs: index.
pages_listA
Returns the full list of pages in the current multi-page document with titles, indices, dimensions, and active status.
Inputs: no named parameters.
pages_switchA
Switches the active canvas view to a specific page index or page ID so the AI agent can design on that page.
Inputs: index, pageId.
pdf_export_catalogA
Compiles all pages (or active page) into a print-ready, high-resolution multi-page PDF document and triggers client-side download.
Inputs: scale, allPages, filename.
project_export_downloadB
Triggers the browser to download the current design as PNG, JPEG, SVG, or project JSON.
Inputs: scale, format, quality, filename.
project_save_localB
Saves the current design project state to the browser local storage / cloud session.
Inputs: no named parameters.
project_set_titleB
Renames the active project title displayed in the top navbar and export filename.
Inputs: title (required).
shape_addA
Adds a geometric shape to the canvas (rect, rrect, circle, triangle, star, heart). Supports fills, strokes, corner radius, and opacity.
Inputs: shapeType (required), x, y, width, +6 more.
shape_update_styleC
Modifies the fill, stroke, corner radius, or opacity of an existing shape layer.
Inputs: layerId (required), opacity, fillColor, strokeColor, +2 more.
sticker_addB
Alias for svg_add. Adds a sticker or badge vector element to the canvas.
Inputs: key (required), x, y, scale.
studio_batch_actionsB
Executes a sequential batch of multiple canvas actions in one round-trip and returns the final visual screenshot and layer state for rapid design creation.
Inputs: actions (required).
studio_check_statusA
Checks whether Tuval Studio is open in the browser and connected to the local WebSocket bridge. If not connected, instructs the AI agent to open https://tuval.site/editor in the user's browser.
Inputs: no named parameters.
studio_export_fileB
Exports the current graphic to a local image file on the host machine (PNG, JPG, WebP, SVG) with customizable resolution scaling (1x, 2x Retina, 3x Print 300DPI, 4x Ultra HD).
Inputs: outputPath (required), scale, format, quality.
studio_get_viewA
Captures the current live canvas graphic and layer hierarchy. Returns an image for visual inspection and critique by vision AI models.
Inputs: scale.
svg_addB
Adds an SVG vector graphic, sticker, or badge to the canvas. Key can be from built-in library ("verified-badge", "sale-badge", "star-burst", "cyber-cross", "arrow-swirl", "sparkle-star", "crown-vip", "fire-flame", "quote-marks", "heart-glow", "abstract-blob") or raw SVG string.
Inputs: x, y, key, svg, +3 more.
templates_listB
Lists all pre-designed templates across all professional design styles (Bento Grid, Corporate, Branding, Luxury, Minimalist, AI & Tech, Retro 80s, Y2K Rave, Bauhaus, Neobrutalism, Art Deco, Memphis Pop).
Inputs: no named parameters.
text_addC
Adds a typography text layer to the canvas with styling (fonts: Sora, Inter, Playfair Display, Montserrat, Poppins, Space Grotesk, Roboto, etc.).
Inputs: text (required), x, y, type, +7 more.
text_update_styleC
Updates typography properties of an existing text layer by its layerId.
Inputs: layerId (required), text, align, color, +5 more.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelogโฆ
TDQS
Scored across 53 tools
The set is mostly distinct by domain and action, but there is an explicit alias (svg_add/sticker_add) and several overlapping background, asset, and export tools that could cause misselection. Descriptions help, but the sheer number of tools increases ambiguity.
Names are overwhelmingly snake_case with domain prefixes (canvas_, layer_, project_, etc.), which is predictable. Minor deviations exist: layers_get_all breaks the layer_ pattern, elements_search vs element_add_stock mix plural/singular, and sticker_add is an alias.
53 tools far exceeds the 3-15 well-scoped range and even the 25+ 'too many' threshold, representing an extreme mismatch for a single MCP server. Many tools could be consolidated, such as background setters, asset listers, and export variants.
The surface covers a wide range of design operations (canvas, layers, pages, assets, text, shapes, export, history), but has notable gaps: no tool to resize or scale a layer, no way to edit existing text content (only style), and no project load/import. These missing operations will likely cause agent dead ends.
Related MCP Connectors
Give your AI agents a design superpower. Generate, edit, and publish publication-grade decks, reports, landing pages, resumes, and marketing visuals directly within your agent workflow. Delivering frontier-level design quality at 3ร the speed and 53ร lower cost -from conversational prompt to live link or vector PDF in minutes.
Agent-Native design tool - create and edit visual designs with agent assistance
Build and run visual creative-production workflows from your AI agent.
Design canvas for AI assets: generate images, video, code and screens, and edit canvases.
Related MCP Servers
AlicenseAqualityCmaintenanceEnables AI agents to generate on-brand visuals from ideas, URLs, documents, or PDFs in over 100 formats and 150+ languages, with consistent brand kits.1627 npm1MIT- FlicenseNot gradedqualityDmaintenanceEnables AI agents to collaboratively design and edit web pages on an infinite canvas using real HTML/CSS, with tools for reading, mutating, and exporting documents.1-
- AlicenseNot gradedqualityCmaintenanceEnables AI agents to act as graphic designers by generating images and social-ready visuals, with brand DNA memory, genre-aware styles, 35+ platform presets, self-critique, and version control.MIT
- AlicenseNot gradedqualityBmaintenanceEnables AI agents to convert natural-language briefs into structured visual specs, deterministically render typography and layouts, and run constraint-aware visual QA and targeted repair for image generation.39 npmApache 2.0