MCP server
Connect any AI assistant to Artimic and design by prompt. The MCP server turns Claude, Codex, Cursor, opencode, and friends into a hands-free design partner that produces real, editable .img projects.
Artimic ships a built-in MCP server — an open-protocol bridge that lets any AI assistant (Claude Code, OpenAI Codex, Cursor, opencode, Continue, Gemini CLI, Cline, and others) drive the editor on your behalf. You stay in your assistant's chat; it produces real, editable .img projects you can open in the editor and refine by hand.
#What this unlocks
Once connected, you can ask your AI assistant things like:
- "Make me a YouTube thumbnail with the title 'Design with AI', a purple gradient background, and a sparkles icon."
- "Open the podcast cover starter template, change the show name to 'Field Notes', and save it as
cover.img." - "Generate a 1080×1920 Instagram story with three product cards stacked vertically — black background, gold accent."
Behind the scenes, the assistant calls the MCP server, the server compiles your description into a valid .img design file, and you get the bytes back. Open the file at the editor and every layer is editable — text, shapes, icons, gradients, effects.
#Why this matters
| Before | With MCP |
|---|---|
| Open editor, click around, position layers manually | Describe the design in plain English; get an editable starting point in seconds |
| Copy-paste hex codes between tabs | Assistant looks up the right preset (Instagram post, YouTube thumbnail, business card, …) |
| Re-do the same thumbnail layout every week | Remix a curated starter template — assistant swaps text/colors and saves a new .img |
| AI generators give you a flat PNG you can't edit | You get a real layered project. Refine in the editor, export to PNG/JPEG/SVG |
The result is a design partner, not a black-box image generator. The assistant gets you 80% there in seconds; the editor lets you take it the rest of the way with no fidelity loss.
#How it works (one paragraph)
The MCP server is a small JSON-RPC endpoint at https://artimic.com/api/mcp. Your AI assistant connects to it once (a one-line config in your client) and gets access to 8 tools: discovering canvas presets, fonts, and curated icons; listing 6+ starter templates; compiling design intents into editor documents; rendering them to .img files; and applying targeted edits to existing designs. The protocol is the Model Context Protocol — an open standard backed by Anthropic and supported by every major AI coding tool.
You don't need to know the protocol. Your assistant handles it. You just tell the assistant what you want.
#What's next
- Connect your AI assistant — one-line setup for Claude Code, Codex, Cursor, opencode, Continue, Cline, Gemini CLI, and any other MCP-compatible client.
- What your assistant can do — tour of all 8 tools, with example prompts you can copy-paste.
- Troubleshooting — common warnings, error codes, and fallbacks if your client doesn't speak MCP yet.
If you're an LLM or building automation directly against the API, jump to For AI agents and The .img format.