Image MCP Server

SceneOf Image MCP Server

Give your AI assistant real visual data

The SceneOf MCP server connects Claude, Cursor, and any MCP-compatible agent to our visual generation APIs. Ask in plain English; get real backgrounds, localization, and edits back.

Connect your assistant

6 APIs exposed
$0 extra cost for MCP connection
Uses existing account credits

The Problem

Your AI is smart. Its visual context is lacking.

LLMs will happily invent details and "generate" images with no connection to your actual products. Grounding them in real localized background APIs changes everything.

Hallucinated cultures

Ask a bare LLM to localize an image for the Middle East and you'll get stereotypical cliches or culturally inappropriate backgrounds.

→ SceneOf MCP returns culturally accurate generated backgrounds your AI can apply.

Copy-paste design loops

Exporting from design tools, removing backgrounds, prompting again, and re-formatting is slower than doing it by hand.

→ The assistant edits the image itself, mid-conversation.

Complex automations

Every custom visual workflow used to mean API integration work before you saw any value.

→ One config line connects Claude or Cursor — no code.

Get Started With SceneOf MCP

First, generate your secure access token. Then, choose your AI client to configure the connection.

New connection

Create an access token

Use a name you will recognize later. The token is displayed once after creation.

Log in to your SceneOf account to generate a secure access token for your AI assistant.

Log in to create token

Account access

Your connections

Sign in to view connections

Active and revoked tokens will be listed here after logging in.

Client recipes

Configure Quick Setup

Replace the token placeholder with the credential shown after creation.

Copy the prompt block below.
Paste it into Claude, Cursor, Codex, or any other AI assistant that can run commands.
The AI will fetch the connection guide and configure the MCP server for your specific client.
Have your SceneOf MCP token ready — it is shown once immediately after creation above.

Paste into Claude, Cursor, Codex, or any AI assistant

Connect my AI assistant to the SceneOf image API via MCP.

Connection details:
  Endpoint:  https://sceneof.com/mcp
  Transport: Streamable HTTP
  Auth:      Authorization: Bearer YOUR_SCENEOF_MCP_TOKEN

Full connection guide (read before starting):
  https://sceneof.com/mcp/docs/connect.md

Instructions:
1. Fetch and read the guide at the URL above.
2. Identify which MCP client I am running.
3. Follow the exact configuration steps for my client from the guide.
4. If your client requires a restart or reload to pick up new tools, instruct me clearly before attempting to verify.
5. Confirm SceneOf image tools are available after the connection is active.

Replace YOUR_SCENEOF_MCP_TOKEN with the token shown after creation.

Configure Cursor

Replace the token placeholder with the credential shown after creation.

In Cursor, open Settings → Tools & Integrations → MCP Tools.
Click Add Custom MCP, paste the configuration from the Config file tab, then save it.
Reopen the chat and confirm SceneOf tools appear under Available Tools.

Use ~/.cursor/mcp.json for every project, or .cursor/mcp.json for the current project only.
Replace YOUR_TOKEN with the token shown after creation. Do not commit the token.

~/.cursor/mcp.json or .cursor/mcp.json

{
  "mcpServers": {
    "sceneof": {
      "url": "https://sceneof.com/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_TOKEN"
      }
    }
  }
}

Configure Claude Code

Replace the token placeholder with the credential shown after creation.

Set SCENEOF_MCP_TOKEN in your shell to the token shown after creation.
Run the command from the Command tab in a terminal.
Run claude mcp list and confirm sceneof is connected.

The command adds SceneOf to Claude Code without editing JSON by hand.

Terminal command

claude mcp add --transport http sceneof https://sceneof.com/mcp --header "Authorization: Bearer $SCENEOF_MCP_TOKEN"

Configure Codex

Replace the token placeholder with the credential shown after creation.

Add this block to ~/.codex/config.toml, or to .codex/config.toml in a trusted project.
Set SCENEOF_MCP_TOKEN in the environment, restart Codex, then use /mcp to verify the connection.

~/.codex/config.toml or .codex/config.toml

[mcp_servers.sceneof]
url = "https://sceneof.com/mcp"
bearer_token_env_var = "SCENEOF_MCP_TOKEN"

Configure OpenCode

Replace the token placeholder with the credential shown after creation.

Merge this server into your opencode.jsonc configuration.
Set SCENEOF_MCP_TOKEN in the environment, restart OpenCode, and verify the SceneOf tools are listed.

opencode.jsonc

{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "servers": {
      "sceneof": {
        "type": "remote",
        "url": "https://sceneof.com/mcp",
        "oauth": false,
        "headers": {
          "Authorization": "Bearer {env:SCENEOF_MCP_TOKEN}"
        }
      }
    }
  }
}

Configure Claude Agent SDK

Replace the token placeholder with the credential shown after creation.

Pass these options when constructing the SceneOf MCP server entry in your agent.
Replace <token> at runtime from secret storage; never commit it in source code.

SDK options

{
  "type": "http",
  "url": "https://sceneof.com/mcp",
  "headers": {
    "Authorization": "Bearer <token>"
  },
  "allowedTools": [
    "mcp__sceneof__*"
  ]
}

Configure Hermes Agent

Replace the token placeholder with the credential shown after creation.

Store SCENEOF_MCP_TOKEN in ~/.hermes/.env, then merge this block into ~/.hermes/config.yaml.
Run /reload-mcp in Hermes and confirm the SceneOf tools are available.

~/.hermes/config.yaml

mcp_servers:
  sceneof:
    url: "https://sceneof.com/mcp"
    headers:
      Authorization: "Bearer ${SCENEOF_MCP_TOKEN}"

Configure OpenClaw

Replace the token placeholder with the credential shown after creation.

Set SCENEOF_MCP_TOKEN in OpenClaw's environment and merge this block under mcp.servers.
Restart OpenClaw and verify the SceneOf tools are available. Do not commit the token.

OpenClaw mcp.servers config

{
  "mcp": {
    "servers": {
      "sceneof": {
        "url": "https://sceneof.com/mcp",
        "transport": "streamable-http",
        "headers": {
          "Authorization": "Bearer ${SCENEOF_MCP_TOKEN}"
        },
        "supportsParallelToolCalls": true
      }
    }
  }
}

Configure Any MCP Client

Replace the token placeholder with the credential shown after creation.

Use this only when the client supports Streamable HTTP and a custom Authorization header.
Keep the bearer token in secret storage and send it on every request.

Streamable HTTP

{
  "url": "https://sceneof.com/mcp",
  "transport": "streamable-http",
  "headers": {
    "Authorization": "Bearer <token>"
  }
}

What's inside

Visual APIs exposed

A complete suite of Image generation and manipulation APIs, available directly to your MCP-compatible agent.

Generate Backgrounds

Create culturally specific, high-quality backgrounds for product images using advanced diffusion models.

Image Translation

Translate embedded text in images while preserving styling, typography, and background context.

Remove Background

Instantly extract subjects from images with pixel-perfect edge detection.

Edit Image

Apply smart edits, compositing, and visual adjustments via natural language prompts.

Who it's for

Who plugs visual workflows into their AI

Ecommerce Teams

Automate generating hundreds of localized product backgrounds for international storefronts directly from your internal AI tools.

Performance Marketers

Ask Claude to ideate ad creative variations and instantly generate the image assets without opening Photoshop.

AI Builders

Give your custom agents the ability to perceive and manipulate images as natively as they handle text.

Pricing

Shared credits. No connection fees.

Connecting your agent via MCP is completely free. You only use standard SceneOf credits when the agent successfully generates or edits an image, exactly as if you used the web UI.

View pricing & credits