dimension-mcp
This server does not currently have a deployable release on Glama. Follow the repository's setup instructions to install and configure it for your MCP client, including any required credentials and local dependencies.
dimension-mcp
Let Claude Code, Codex, or any MCP client drive Adobe Dimension. Build product mockups and 3D scenes by describing them: add models, set materials and colors, put labels and logos on products, match a background photo, frame the camera, render, and save.
Unofficial, community project. Not affiliated with or endorsed by Adobe. "Adobe" and "Adobe Dimension" are trademarks of Adobe Inc.
Why
Adobe Dimension has no scripting API and no official AI connector. But it is an Electron app, and Electron apps can be started with a local debug port. Through that port this server reaches Dimension's own controllers (scene graph, materials, decals, cameras, lights, render, documents), so every edit goes through the app's normal code paths and stays undoable with Cmd/Ctrl+Z.
Related MCP server: Illustrator AI MCP Server
What it can do
Tool | What it does |
| Start Dimension in control mode, or connect if it already is |
| Current document and the scene tree with paths |
| See the result: render-camera snapshot, or the whole window |
| Search the 570+ starter assets and add them, or import OBJ / FBX / GLB / GLTF / STL / SKP / DN |
| Position, rotation, scale, size (cm, degrees) |
| Read or set base color, roughness, metallic, opacity, emission, or a texture image |
| Put a label, logo, or artwork on a mesh |
| Use a photo as the background and run Match Image (perspective, lighting, canvas) |
| Preset views (FRONT, TOP, FRONT_RIGHT_TOP, …) and framing |
| Final photoreal render to PNG |
| Info, save, open, new |
| Run JavaScript inside Dimension — full control for anything else |
| Search Dimension's own source (unpacked locally from your install) to learn how a feature works |
It pairs well with Illustrator, Photoshop, and Blender MCP servers: design a label in Illustrator, build a missing shape in Blender, then stage and render it in Dimension.
Requirements
Adobe Dimension 4.x (tested on 4.1.9, macOS)
Node.js 20 or newer
An MCP client: Claude Code, Codex, Claude Desktop, Cursor, …
Install
The server runs straight from this GitHub repository with npx: nothing to clone or build. Node.js must be installed (node -v should print 20 or higher).
macOS | Windows | |
Claude Code | tested | expected to work, not tested yet |
Codex | expected to work | expected to work, not tested yet |
Claude Code: macOS
claude mcp add dimension --scope user -- npx -y github:fatoomdes-lang/dimension-mcpClaude Code: Windows (PowerShell or Command Prompt)
On Windows, npx has to be started through cmd /c:
claude mcp add dimension --scope user -- cmd /c npx -y github:fatoomdes-lang/dimension-mcpCodex: macOS
codex mcp add dimension -- npx -y github:fatoomdes-lang/dimension-mcpCodex: Windows
codex mcp add dimension -- cmd /c npx -y github:fatoomdes-lang/dimension-mcpOr add it to ~/.codex/config.toml (on Windows use command = "cmd" and put "/c", "npx" first in args):
[mcp_servers.dimension]
command = "npx"
args = ["-y", "github:fatoomdes-lang/dimension-mcp"]Other clients (Claude Desktop, Cursor, …)
{
"mcpServers": {
"dimension": { "command": "npx", "args": ["-y", "github:fatoomdes-lang/dimension-mcp"] }
}
}Optional: the skill (Claude Code)
It teaches Claude the best workflow in Dimension. Codex reads AGENTS.md from this repo instead.
macOS:
mkdir -p ~/.claude/skills/dimension && curl -fsSL https://raw.githubusercontent.com/fatoomdes-lang/dimension-mcp/main/skills/dimension/SKILL.md -o ~/.claude/skills/dimension/SKILL.mdWindows (PowerShell):
New-Item -ItemType Directory -Force "$HOME\.claude\skills\dimension" | Out-Null; Invoke-WebRequest https://raw.githubusercontent.com/fatoomdes-lang/dimension-mcp/main/skills/dimension/SKILL.md -OutFile "$HOME\.claude\skills\dimension\SKILL.md"Start a new session after adding it — MCP tools load when a session starts.
Use
Ask in plain words, for example:
Open Dimension. Add a cream tube, make it navy and glossy, put ~/Desktop/logo.png on the body, use ~/Desktop/cafe.jpg as the background and match it, then render to the Desktop.
The agent calls dimension_launch first. If Dimension is already open the normal way, it asks you to save, then restarts it in control mode.
How it works
Dimension is launched with
--remote-debugging-port=9222(localhost only).The server connects over the Chrome DevTools Protocol to Dimension's main window.
It injects a small helper library (
src/prelude.browser.js) that wraps Dimension's controllers:window.debugController.appis the app's own controller tree.Dimension runs Chrome 49, so code sent by the agent is lowered to ES2015 with esbuild before it runs.
Security
While Dimension runs in control mode, any program on the same computer can control it through the debug port (it listens on 127.0.0.1 only; nothing is reachable from the network). Quit Dimension when you are done; a normal launch turns the port off.
dimension_source unpacks Dimension's source map from your own installation into ~/.dimension-mcp/source so the agent can read it. No Adobe code is included in this repository.
Limitations
Undocumented internal APIs: they may change if Adobe updates Dimension (development is paused, so this is unlikely).
One debug connection at a time: only one agent session can control Dimension at once.
Windows should work the same way but is not tested yet. Set
DIMENSION_APPto the path ofAdobe Dimension.exeif it is not in the default location. Reports welcome.DIMENSION_DEBUG_PORTchanges the port (default 9222).
Development
git clone https://github.com/fatoomdes-lang/dimension-mcp.git
cd dimension-mcp && npm install
npm run smoke # read-only checks against a running DimensionPoint your client at node /absolute/path/to/dimension-mcp/src/index.js while developing.
License
MIT © Fatimah Almehmadi
Related MCP Connectors
Make games with AI: one MCP server that connects your agent to Unity, Unreal and Godot editors.
Generate game-ready 3D models, textures, and audio from natural language, over MCP.
Brand-safe MCP for AI agents to create editable, on-brand graphics and automate variants.
Generate on-brand images from your AI agent: design, edit, and render templates over MCP.
Related MCP Servers
- AlicenseAqualityDmaintenanceA professional MCP server that lets AI assistants directly control Adobe Photoshop for document creation, layer management, filters, adjustments, shapes, text, and design intelligence like palette generation and font pairing.322MIT
- FlicenseNot gradedqualityDmaintenanceAn MCP server that lets AI assistants control Adobe Illustrator locally—running ExtendScript, capturing canvas screenshots, exporting artwork, and optionally remembering techniques—all without network activity.-
- AlicenseAqualityDmaintenanceAn MCP server that allows AI agents to control Adobe After Effects, enabling project inspection, composition creation, layer addition, file import, ExtendScript execution, and rendering via aerender.11MIT
- AlicenseNot gradedqualityDmaintenanceAn MCP server that enables AI assistants to programmatically control Adobe Photoshop through natural language, supporting design, image manipulation, and workflow automation.5,091 npmMIT