Skip to content
MCP server

Modellix Agent Canvas

Local infinite canvas for AI image generation, editing, HTML drafts and slide decks

Runs locallyActively maintained

The repository was last updated 35 days ago and is published under MIT.

Install
Pricing
Paid
Platforms
CLI, Web, macOS
Runs
Locally
GitHub stars
0
Last commit
1 month ago
Language
HTML

Similar MCP servers

About Modellix Agent Canvas

Modellix Agent Canvas is a paid MCP server for design and media. It runs locally on your machine and is started by your AI client.

Modellix Agent Canvas runs on your own machine as a workspace-bound stdio MCP plugin. It embeds an Excalidraw infinite canvas where an agent and a person can generate and edit images through Modellix models, draft HTML with a sandboxed live preview, build presentations and keep everything in the project folder. Hosts that support MCP Apps show the whole canvas inline; others open a short-lived page on localhost.

Spending is gated. Preparing an image task is free and returns the chosen model, the routing reason, the effective specification, quantity and any available price quote; a separate one-time confirmation is required to submit, duplicates are deduplicated locally and an unknown submission is never retried automatically. Task IDs and results are written to an append-only ledger so work survives a closed host. The README warns not to confirm a task with no quote if you have a fixed budget.

It supports multi-page projects, image placeholders, up to ten ordered references, PNG and SVG export at up to 4x, slide-deck PNG sequences and project backups. The interface is in English, Chinese and Japanese. You need Node.js and a Modellix API key, stored in the system credential store.

Install

Run in a terminal

claude mcp add agent-canvas -- npx -y @modellix/agent-canvas

Add to claude_desktop_config.json

{
  "mcpServers": {
    "agent-canvas": {
      "command": "npx",
      "args": [
        "-y",
        "@modellix/agent-canvas"
      ]
    }
  }
}
Install in CursorOpens Cursor, which asks you to confirm. Or add it by hand:

Add to ~/.cursor/mcp.json

{
  "mcpServers": {
    "agent-canvas": {
      "command": "npx",
      "args": [
        "-y",
        "@modellix/agent-canvas"
      ]
    }
  }
}
Install in VS CodeOpens VS Code, which asks you to confirm. Or add it by hand:

Add to .vscode/mcp.json

{
  "servers": {
    "agent-canvas": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "@modellix/agent-canvas"
      ]
    }
  }
}

Add to ~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "agent-canvas": {
      "command": "npx",
      "args": [
        "-y",
        "@modellix/agent-canvas"
      ]
    }
  }
}

Add to your client's MCP settings (most use this shape)

{
  "mcpServers": {
    "agent-canvas": {
      "command": "npx",
      "args": [
        "-y",
        "@modellix/agent-canvas"
      ]
    }
  }
}

Using several servers? Build one config file with the MCP config generator.

Tools this server provides (3)

Taken from the project documentation; the running server may differ.

get_modellix_canvas_status
Reports canvas status.
open_modellix_canvas
Opens the canvas for a workspace.
prepare_modellix_image_task
Prepares an image task and shows model, route and price disclosure.

What it does

Excalidraw canvas
Infinite canvas with shapes, images, pages and undo.
Image generation and editing
Text-to-image and reference-based edits with 1 to 4 outputs.
Cost preview
Prepares a task for free and shows model and price before submitting.
Durable tasks
Keeps a local ledger so work resumes after a restart.
HTML drafts
Edits source beside a sandboxed preview and captures it.
Presentations
Builds slide decks, plays them and exports PNG sequences.

Good for

  • Edit an image with a chosen reference
  • Check the cost of an image task before paying
  • Prototype an HTML layout and capture it as an image
  • Build a slide deck from generated visuals
  • Back up and restore a visual project

Pricing

Paid

Free local app; image generation billed through your Modellix API key (prices not listed)

Modellix Agent Canvas FAQ

Does it need a cloud service?

No. The canvas runs locally; image generation calls the Modellix API with your key.

Can it spend money without asking?

No. Preparation is free and submission needs an explicit confirmation.

Which hosts work?

Codex, Cursor, Claude Code, OpenCode and other local stdio MCP hosts.

Where is the API key stored?

In the system credential store, not in chat, URLs or project files.

Is Modellix Agent Canvas free?

No. Modellix Agent Canvas is a paid product.

How do I add Modellix Agent Canvas to Claude Code?

Run this command in your terminal: claude mcp add agent-canvas -- npx -y @modellix/agent-canvas

What are the best alternatives to Modellix Agent Canvas?

The closest alternatives to Modellix Agent Canvas are Banana Art Lab, Maket, next-ai-draw-io, ExceedLoop Sketchpad and Vmotif.

Where this comes from

  • Description: written from the project's README.
  • Install steps: from the official MCP Registry entry.

How AIPoweredBot checks servers