The short answer: not on its own — but yes with one connector. Claude is a text-only model. Ask it to draw something and it will politely explain that it can’t. But Claude supports the Model Context Protocol (MCP), which means you can hand it an image generator. Connect the AI Box MCP and Claude can generate images with GPT Image, DALL·E 3, Flux, Ideogram, Stable Diffusion and more — directly inside your conversation, on both the free and Pro plans. Setup takes about 60 seconds and there are no API keys involved.
Why can’t Claude make images natively?
Anthropic built Claude as a text (and vision input) model — it can look at images you upload, but it has no image generation model behind it. That’s a deliberate product choice, and it’s why every “Claude image generation” solution works the same way: an external image model, wired into Claude through MCP, Anthropic’s open standard for giving AI assistants tools.
How to add image generation to Claude (60 seconds)
- Open Claude’s connector settings. In Claude (web or desktop), click your profile → Settings → Customize → Connectors.
- Add a custom connector. Click +, choose Add custom connector, name it AI Box, and paste this URL:
https://mcp.aibox.ai/mcp/ - Authorize. Claude pops over to AI Box for a one-click OAuth approval. No JSON, no API keys.
- Ask for an image. Open a new chat and try: “Generate a 16:9 blog cover image about remote work trends — editorial photography style, warm tones.” The finished image lands right in the conversation.
Full walkthrough with screenshots: Set up the AI Box MCP in Claude.
Which image models can Claude use?
Through AI Box, Claude gets every top image model under one subscription: GPT Image, DALL·E 3, Flux, Ideogram (best-in-class for text inside images), Stable Diffusion, DreamShaper, and HiDream. Describe what you want in plain English and Claude routes it to a fitting model — or name the model you want (“use Flux for this one”).
What people actually make with it
- Blog covers and article art while Claude writes the post
- Product shots and ad-creative variations
- Logo and brand concepts (generate three options, iterate in the same chat)
- Social graphics sized per platform
- UI mockup imagery while vibe-coding
Because the writing and the visuals happen in one conversation, the iteration loop is fast: Claude drafts the landing page copy, generates the hero image, you say “warmer, and put the product on the left,” and it regenerates.
What does it cost?
Claude’s side is free — custom connectors work on Claude free and Pro. The generation itself runs on your AI Box plan, which starts at $8.99/month and includes all 80+ models (image, video, audio, and text). There’s nothing extra to pay for the MCP itself.
Beyond images: video and audio too
The same one-URL connection also gives Claude video generation and audio generation — text-to-video clips, voiceovers with ElevenLabs and OpenAI TTS, even full multi-step workflows you build yourself in the no-code Builder. One connector, every modality.
FAQ
Does this work in the Claude mobile app? Connectors you add on the web carry over to your account; generation works anywhere your connectors are available.
Where do my images go? They appear in the chat and are also saved to your AI Box media storage, so nothing is lost if a conversation is deleted.
Does it work with Claude Code or Cursor? Yes — the same MCP URL works in every MCP-compatible client. See the Cursor guide.
Is there a way to do this without any third-party service? You can self-host an open-source image MCP server and bring your own model API keys — more control, but you’re managing keys, hosting, and per-provider billing yourself. AI Box exists to skip all of that.
Ready to try it? Grab the URL — https://mcp.aibox.ai/mcp/ — and follow the 60-second setup guide.