What Claude can make on its own
According to Anthropic's help center, Claude can build diagrams, charts and interactive visuals in the conversation, made with HTML and SVG, so you can ask follow-up questions and have Claude adjust them. Anthropic lists this as a beta on Claude web and desktop. Claude can also read and analyze images you upload.
What it does not make is a picture: a photo, an illustration, a product shot or a poster rendered by an image model. That is what a connector adds.
Which image models Claude can use through MakeMidas
MakeMidas lists image models from Google (Nano Banana), OpenAI (GPT Image), ByteDance (Seedream), Black Forest Labs (Flux), xAI (Grok Imagine) and Alibaba (Qwen Image). Name one in your request ("with Nano Banana Pro"), or leave it out and MakeMidas uses the cheapest model your account may use. Claude can also run one prompt on two to four models and show them side by side.
Each image is charged to your MakeMidas credits at the same price as on the website; the pricing page lists every model. Ask Claude to estimate a request first if you want the price before anything is charged.
Editing a picture you already have
Most of these models take reference images, so Claude can edit a photo, restyle it or combine several into one. Claude does not pass files you attach to connectors. To start from a picture, paste a public link to it, or ask Claude for a MakeMidas upload link, which opens MakeMidas in your browser to add the picture there.
In Claude Code
In Claude Code, run claude mcp add --transport http makemidas https://makemidas.com/api/mcp, then type /mcp to sign in. Claude Code can also run the MakeMidas CLI, which saves results as files in your project; the agent skills teach it the options.