Recraft now supports the Model Context Protocol, letting you generate images directly through Claude, Cursor, and other AI agents without switching between tools.
When AI models operate in the wild beyond chat windows and playgrounds, the question isn't "Can the model generate something?" It's "Can the model work within a real system?"
This is where context becomes everything: tools, data, user intent, and workflow logic. Historically, integrating those pieces has been clunky — messy snippets of prompt hacks, hand-coded APIs, or manual copy-paste workarounds. The approach doesn't scale. It doesn't generalize. And it certainly doesn't enable the kind of modular, intelligent behavior that creative workflows demand.
That's why Recraft is now live with MCP (Model Context Protocol), an open standard designed to make tools like ours deeply accessible to the next generation of AI agents and chat interfaces.
At a technical level, MCP defines a standardized interface for exposing three key things to AI models:
It's not just a protocol, it's a bridge between AI models and the real workflows people care about.
Instead of cramming everything into one giant prompt and hoping for the best, MCP treats context like live infrastructure. Tools are declarative. Prompts are structured. Resources are discoverable. And any model that integrates with MCP, like Claude or Cursor, can understand and use those tools instantly.
Our goal is straightforward: make it possible for people to use Recraft's creative AI inside the emerging world of intelligent agents.
With Recraft supporting the Model Context Protocol, you can connect Recraft's image generation engine directly to Claude, Cursor, and other MCP-capable interfaces. This unlocks new ways of interacting with Recraft — via chat — where AI can ideate, style, and generate visuals through Recraft tools in real time.
Recraft remains a standalone product, but with MCP, we’re extending our tools into the agent ecosystem so users can access Recraft wherever they work or experiment.
The most exciting part is this doesn’t just enable access, it enables powerful conversational co-creation.
With MCP, AI agents like Claude don’t just suggest ideas; they can execute them directly through Recraft’s tools, turning the design process into a continuous, integrated dialogue between you, the LLM, and Recraft.
Imagine you're solving a design task. Instead of bouncing between tabs or tweaking prompts manually, you could:
The entire design process becomes conversational, iterative, and integrated. That's what MCP makes possible.
If you've ever built custom glue code to get an LLM to call a tool, you know the pain. You're writing brittle prompt wrappers, dealing with inconsistent context injection, and reinventing the same logic over and over.
With MCP, you write one server. Claude, Cursor, and any other client that supports MCP just works. Your Recraft integration becomes discoverable, modular, and scalable. You stop thinking about "how do I format this prompt" and start designing systems.
For developers building with Recraft, this means faster experimentation, clearer contracts between model and tool, and a much smoother path to intelligent workflows.
Recraft's MCP server is functional, open, and available for anyone to use. Just head over to Recraft MCP on Github to get started. Enjoy!
