MarkItUp MCP server — AI image annotation and marketing-visual generation for Claude Code, Cursor, Claude Desktop, ChatGPT, and other MCP clients
markitup-mcp-server
以下摘录自该仓库的 README(片段) · 在 GitHub 读完整 README →
MCP (Model Context Protocol) server for MarkItUp — AI-powered image annotation and marketing-visual generation. Lets Claude Code, Claude Desktop, Cursor, ChatGPT-with-MCP, and other MCP-compatible clients call MarkItUp directly.
Five tools available: markitup_credit_balance, markitup_generate, markitup_regen, markitup_extend, markitup_remove_background.
Prerequisites
1. A MarkItUp account — sign up at https://markitup.app.
2. A MarkItUp API key — generate one at https://markitup.app/dashboard.
3. Node.js 18 or newer (only if installing locally; npx doesn't require a local Node.js if your MCP client bundles one).
Configure in Claude Code
Add to your MCP config (~/.claude/mcp.json or project-scoped .mcp.json):
Configure in Claude Desktop
Edit ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):
Configure in Cursor
Cursor uses the same JSON shape. Settings → MCP → Add Server:
Tools
markitup_credit_balance
Returns the current credit balance and subscription status.
No arguments. Use it to verify your API key works and to check credits before calling generation tools.
markitup_generate
Generates 3 polished marketing-visual variations from a screenshot. Costs 1 credit.
| Argument | Type | Required | Description |
|---|---|---|---|
| description | string | yes | What the image shows and what to highlight |
| template_id | string | yes | One of glassmorphic, clean_minimal, bold_marketing, dark_professional, documentation (and more) |
| image_url | string | one-of | Public HTTPS URL of the source image |
| image_base64 | string | one-of | Base64-encoded image bytes (no data: prefix) |
| image_mime_type | string | no | MIME type when supplying image_base64. Default: image/png |
| aspect_ratio | string | no | One of 1:1, 3:2, 2:3, 3:4, 4:3, 4:5, 5:4, 9:16, 16:9, 21:9 |
| image_size | string | no | 1K, 2K, or 4K |
Returns: the marketing copy (headline, sub-headline) plus 3 generated image variations as base64 image content blocks. The full structured response is also available under structuredContent.
**Example prompt (in Claude Code):**
> Use markitup_generate with the screenshot at https://example.com/dashboard.png — highlight the export button and pitch it for a marketing landing page using the bold_marketing template.
markitup_regen
Regenerate one variation from a previous markitup_generate call. Costs 1 credit by default (charge_credit: false to skip).
The LLM must pass back the text_analysis object from the previous generate's structuredContent.text so the regenerated visual stays on-brand with the same headline / sub-headline. variation_index selects which slot (0–2).
markitup_extend
AI outpaint — extend an image to a larger canvas / different aspect ratio. Costs 1 credit.
| Argument | Type | Required | Description |
|---|---|---|---|
| image_url / image_base64 | string | one-of | Source image |