Back to articles
Tutorial

OpenArt MCP: Generate Images & Videos Directly Inside Claude, ChatGPT, and Cursor

OpenArt just shipped MCP integration, letting you generate professional images and videos without leaving your AI agent. One URL, zero API keys, instant access to all OpenArt models. Here's how to set it up and why it changes the game for developers and creators.

8 min read

Stop Context Switching: Generate Content Where You Already Work

The workflow used to look like this:

1. You're in Claude asking for help building something

2. Claude suggests creating visuals, but can't

3. You tab over to OpenArt.ai

4. Generate the image there

5. Download and bring it back

6. Resume your conversation

That's context fragmentation. It breaks flow. It wastes time.

OpenArt MCP just collapsed this entire workflow into a single interface.

---

What is MCP?

MCP stands for Model Context Protocol -- essentially a bridge that lets AI agents (Claude, ChatGPT, Cursor) directly call external tools and services.

OpenArt's MCP integration means:

  • Generate images and videos directly in Claude/ChatGPT without leaving the chat
  • Access all OpenArt models (GPT Image 2, Flux 2 Pro, Seedance 2.0, SDXL, Nano Banana 2) from within your AI agent
  • No API keys to manage -- just OAuth sign-in
  • Automatic project organization -- outputs sync to your OpenArt library
  • Reuse previous generations -- reference past images as style guides
  • It's the difference between "call an external API" and "this tool is part of my workspace."

    ---

    How to Set It Up (5 Minutes)

    For Claude:

    1. Go to Settings → Connectors

    2. Click Add custom connector

    3. Paste this URL: `https://mcp.openart.ai/mcp`

    4. Click Add connector

    5. Sign into your OpenArt account when prompted

    6. Done. Claude now has OpenArt powers.

    For ChatGPT:

    1. Go to Settings → Developer mode

    2. Click Create connector

    3. Enter:

    - Name: OpenArt

    - URL: `https://mcp.openart.ai/mcp`

    4. Sign in to OpenArt when prompted

    5. You're connected

    For Cursor (For Developers):

    1. Open `.cursor/mcp.json` in your project root

    2. Add this configuration:

    {
      "mcpServers": {
        "openart": {
          "command": "npx",
          "args": ["-y", "@openart-ai/mcp-server"]
        }
      }
    }

    3. Restart Cursor

    4. You can now generate images directly in your code editor

    ---

    Practical Examples

    Example 1: Content Creator in Claude

    You: "I'm writing a blog post about AI video generation. Can you generate a hero image showing someone editing video with AI?"

    Claude: Generates the image directly, shows you a preview, saves it to your OpenArt library, and embeds it in the conversation.

    No tabs. No downloads. No friction.

    Example 2: Developer in Cursor

    You're building a landing page and ask Cursor to:

    "Generate 5 hero section background images for a SaaS product dashboard. Style: modern, tech-forward, blue and purple gradient."

    Cursor: Generates all 5 images, saves them to your project, and you can reference them directly in your code.

    Example 3: Marketing Team in ChatGPT

    You need 10 product variations for A/B testing:

    Prompt: "Generate 10 variations of our product (a blue water bottle) in different environments: office, gym, beach, mountain, etc."

    ChatGPT + OpenArt MCP: Generates all 10 in one go, organized in your OpenArt library by project.

    ---

    Why This Matters

    For Developers:

  • Generate placeholder images and backgrounds without leaving your editor
  • Prototype visual designs before building the UI
  • Batch-generate assets for multiple projects simultaneously
  • No need to wrangle APIs or manage credentials
  • For Content Creators:

  • Brainstorm and generate content in natural conversation flow
  • Instantly create multiple variations for testing
  • Stay in your AI workspace instead of context-switching
  • Reuse style references across projects
  • For Teams:

  • Shared project organization across OpenArt
  • Everyone on the team can generate from their preferred AI agent
  • Automatic sync to your OpenArt account
  • No per-user API key management headaches
  • ---

    What You Can Generate

    Since you're connecting to the full OpenArt platform, you get access to:

    Image Models:

  • GPT Image 2 (text to ultra-realistic images)
  • Flux 2 Pro (fast, high-quality generation)
  • SDXL (versatile, great for art styles)
  • Nano Banana 2 (multimodal, text + images in one model)
  • Video Models:

  • Seedance 2.0 (ByteDance's cinematic video generation)
  • Smart Shot (multi-shot cinematic sequences)
  • Video Editing:

  • Replace Background (swap environments)
  • Relight Video (adjust lighting and mood)
  • VFX (mask and transform specific elements)
  • ---

    Limitations

    What MCP can't do (yet):

  • Real-time video editing with frame-by-frame preview
  • Advanced masking and selection tools (those are better in the web UI)
  • Batch operations with complex scheduling
  • Collaborative editing sessions
  • When to use the web UI instead:

    If you need precise control over masking, detailed editing, or are doing heavy-duty project work, OpenArt.ai's web interface is still the go-to.

    MCP is best for quick generation, iteration, and integration into your workflow -- not pixel-perfect production work.

    ---

    Credit Usage

    MCP uses the same credit system as the web interface:

  • Image generation: 10-40 credits per image (varies by model)
  • Video generation: 200-500 credits per video
  • Video editing: 3-6 credits per second
  • No hidden charges. No "MCP premium tier." Same rates, same models, same library.

    ---

    The Real Benefit: Workflow Integration

    Here's what changed for me:

    Before MCP:

  • Open Claude
  • Describe image I want
  • Realize I need to actually generate it
  • Tab to OpenArt.ai
  • Generate image
  • Download it
  • Tab back to Claude
  • Upload image
  • Continue conversation
  • With MCP:

  • Open Claude
  • Describe image
  • Claude generates it
  • Continue conversation
  • That's not a small optimization. That's the difference between "tool I have to context-switch to" and "part of my workspace."

    For developers using Cursor, it's even more dramatic. You can generate assets without ever leaving your editor.

    ---

    Who Should Use This

    Perfect for:

  • Developers prototyping with AI-generated assets
  • Content creators working in ChatGPT/Claude
  • Teams managing multiple content projects
  • Anyone who values workflow efficiency
  • Not ideal for:

  • Production-grade pixel-perfect editing (use web UI)
  • Complex multi-layer masking workflows
  • People who never use AI agents anyway
  • ---

    Getting Started

    The setup is genuinely 5 minutes. Just pick your agent (Claude/ChatGPT/Cursor) and add the MCP URL.

    After that, your next prompt asking for image generation will just... work.

    Try OpenArt MCP Free - Connect any AI agent and start generating immediately. Same 20,000 free credits, new workflow possibilities.

    Found this helpful?Share this article with your network to help others discover useful AI insights.

    Share your thoughts on X or Facebook