One Remote Endpoint
A single Streamable HTTP endpoint connects with Cursor, Claude Code, Codex, Windsurf, and custom agent runtimes.
MODEL CONTEXT PROTOCOL · REMOTE SERVER
Use ImageToLayers from your AI agent
Connect Cursor, Claude Code, Codex, and other MCP-compatible clients. Vectorize graphics, extract transparent layers, and inspect job assets directly from your coding assistant.
Live Remote Endpoint
Streamable HTTPOAuth 2.1
Server URL
https://www.imagetolayers.com/mcp
No API keys exposed · Authenticates with your ImageToLayers loginView client setup guide →
FAST SETUP
Connect your favorite AI agent
Add the ImageToLayers endpoint to your client configuration. When prompted, complete the one-click browser authorization.
~/.cursor/mcp.json
Add to your Cursor MCP settings or workspace configuration file.
{
"mcpServers": {
"imagetolayers": {
"url": "https://www.imagetolayers.com/mcp"
}
}
}After saving, trigger any tool from your agent prompt (e.g. “Generate an SVG of a minimal coffee cup”).
TOOL REFERENCE
Tool reference
Parameters and output for each function. Reuses the identical high-performance pipelines that power the web studio.
Paid · Uses account credits
Text to native vector SVGGenerates clean, scalable, multi-path SVG vector graphics from any natural language prompt. Returns optimized SVG markup and a preview URL.
Parameters
prompt*(string)Detailed text description of the vector graphic to create.
style(string)Optional artistic direction: 'flat', 'minimal', 'line-art', or 'detailed'.
Returns:Job ID + direct SVG XML content & preview image asset
Paid · Uses account credits
Raster image to layered PSD/ZIPDecomposes a flat raster photo or illustration into isolated semantic layers (subject, foreground objects, background, lighting).
Parameters
image_url*(string)Public HTTPS URL of the image to decompose.
export_format('psd' | 'zip')Preferred container format (defaults to 'psd').
Returns:Job ID with background pipeline tracking
Paid · Uses account credits
Outpaint an image to a new ratioExtends an image beyond its original borders, generating new content in the added area to reach a target aspect ratio. Runs asynchronously — poll get_job for the result.
Parameters
image_url*(string)Public HTTPS URL of the image to extend.
aspect_ratio*('original' | '1:1' | '4:5' | '9:16' | '16:9' | '3:2' | '2:3')Target aspect ratio for the extended canvas.
position(string)Where the original sits in the new canvas (e.g. 'center', 'bottom-center'). Defaults to 'center'.
prompt(string)Optional guidance for the generated area. Omit to continue the scene naturally.
request_id*(string)Caller-supplied unique id; retrying the same id returns the original job instead of charging again.
Returns:Job ID; get_job returns a signed PNG download URL on completion
Read-only · Free
Job status & asset download linksPolls or retrieves the completion status and download URLs (PSD, PNG layers, SVG) for any asynchronous processing job.
Parameters
job_id*(string)Unique job identifier returned from paid tool calls.
Returns:Status ('queued' | 'processing' | 'succeeded' | 'failed') and signed asset download URLs
Read-only · Free
Credit balance & account quotasQueries the remaining credit balance, active subscription tier, and monthly quota of the authenticated ImageToLayers account.
Returns:Available balance, plan name, and current billing cycle status
CAPABILITIES
Supported MCP functions
Five tools your agent can call over the remote endpoint. Inspection tools are free; generation tools draw from your credit balance.
- Uses credits
generate_svg_from_textGenerate a clean, multi-path vector SVG from a text prompt.
- Uses credits
convert_image_to_layersDecompose a raster image into isolated semantic layers, exported as PSD or ZIP.
- Uses credits
extend_imageExtend an image beyond its borders (outpainting) to a target aspect ratio.
- Free
get_jobCheck the status of an async job and retrieve signed asset download URLs.
- Free
get_account_usageRead your remaining credit balance and recent usage.
DEVELOPER ARCHITECTURE
Built for security and reliability
Designed from the ground up to keep user credentials safe while delivering production-grade performance.
Zero API Key Leak
Your agent authenticates via Supabase OAuth 2.1 in a standard browser prompt. No raw API keys or tokens are hardcoded into local repos.
Shared Credit Pool
Agent calls draw directly from your existing ImageToLayers credit balance. No separate billing, credit cards, or subscription tiers.
Production Pipeline Parity
All MCP tools run on the identical model cluster and segmentation infrastructure as the official ImageToLayers web application.
WORKFLOW
How the connection works
From configuration to autonomous agent tool calls in under two minutes.
Step 01
Add endpoint to your agent
Copy the remote URL into your client configuration (e.g. Cursor mcp.json or Claude Code CLI).
Step 02
Approve OAuth connection
Your agent initiates a standard browser OAuth handshake. Review requested scopes and approve with one click.
Step 03
Prompt your agent naturally
Ask your agent to vectorize graphics, extract transparent layers, or generate SVGs without leaving your IDE.
Pre-flight balance checks & zero overdraft
Before executing any paid image segmentation or vector generation task, the server verifies that your account has sufficient credits. Requests that exceed your quota are cleanly rejected before incurring any charge.
INTEGRATIONS
Explore the developer ecosystem
Agent Workspace
Prefer an interactive web canvas? Use ImageToLayers’ built-in AI workflow assistant with real-time visual previews and prompt templates.
Open Agent WorkspaceDirect API Access
Building your own commercial product or high-volume backend pipeline? Request direct HTTP REST API access with custom SLAs.
Explore API accessQUESTIONS & ANSWERS
Frequently asked questions
What is the ImageToLayers MCP Server?
It is a Remote Model Context Protocol (MCP) server that lets AI agents in Cursor, Claude Code, Codex, and other environments invoke ImageToLayers tools programmatically. Agents can generate vector SVGs, decompose photos into layered PSDs, and inspect job status directly from your conversation prompt.
What is the Model Context Protocol (MCP)?
MCP is an open standard developed by Anthropic that connects AI assistants and agents to external developer tools, data sources, and APIs through a standardized JSON-RPC interface.
Which clients and IDEs are supported?
Any MCP client that supports Streamable HTTP with OAuth 2.1 authentication, including Cursor, Claude Code CLI, Codex, and Windsurf.
Do I need a separate account or API token?
No. You authenticate with your existing ImageToLayers user account. The client opens a secure OAuth 2.1 window where you approve access, eliminating the need to paste or store raw API tokens on disk.
How are credits charged?
Tool usage consumes standard ImageToLayers credits from your existing balance. Read-only inspection tools like get_job and get_account_usage are completely free.
Can an agent overspend my balance?
No. Pre-flight quota verification runs before any generation job starts. If your balance is insufficient, the request is safely rejected before any action is taken.
How do I revoke access?
You can revoke an agent's access token immediately from your ImageToLayers account dashboard. Once revoked, the agent cannot execute any further operations without re-authorizing.
How does this differ from the Agent Workspace (/agent)?
The Agent Workspace is our interactive web canvas for visual and layer manipulation in the browser. The MCP Server allows external agents running in your local IDE or CLI to call the same backend capabilities directly.
Is /mcp a web page I can view in the browser?
No. The /mcp endpoint is a machine-to-machine JSON-RPC interface. This page (/mcp-server) is the human-readable developer documentation and quickstart guide.
Connect your AI assistant today
Point any MCP-compatible agent at our live endpoint and start transforming images from your prompt bar.