The MarkItUp MCP Server is the MCP (Model Context Protocol) server for MarkItUp, used for AI-powered image annotation and marketing visual generation. It allows Claude Code, Claude Desktop, Cursor, MCP-enabled ChatGPT, and other MCP-compatible clients to call MarkItUp directly.
Purpose The server is aimed at scenarios that need to quickly produce marketing visual assets from screenshots, connecting MarkItUp's image generation and editing capabilities to commonly used AI clients through the MCP protocol.
Main capabilities The server provides five tools:
- markitup_credit_balance: Returns the current credit balance and subscription status. It requires no parameters and can be used to verify whether the API key is valid and to check credits before calling generation-type tools.
- markitup_generate: Generates 3 polished marketing visual variations from a single screenshot, consuming 1 credit. Returns marketing copy (main headline, subheadline) and 3 generated image variations (returned as base64 image content blocks); the full structured response is also available in structuredContent.
- markitup_regen: Regenerates one variation based on the results of a previous markitup_generate call. By default it consumes 1 credit (can be skipped with charge_credit: false). The LLM must pass back the text_analysis object from structuredContent.text of the previous generation result to ensure the regenerated visual maintains brand consistency with the same main headline/subheadline; variation_index is used to select the slot (0–2).
- markitup_extend: AI outpainting, extending an image to a larger canvas or a different aspect ratio, consuming 1 credit.
- markitup_remove_background: Removes the image background using Photoroom's HD model, returning a transparent PNG, consuming 1 credit (free for active Pro/Power subscribers).
How to use Prerequisites: A MarkItUp account (register at https://markitup.app ), a MarkItUp API key (generated at https://markitup.app/dashboard ), and Node.js 18 or higher (only required for local installation; if the MCP client bundles Node.js, using npx does not require local Node.js).
Configure in Claude Code: Add the configuration to the MCP configuration file (~/.claude/mcp.json or the project-level .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 structure; add it via Settings → MCP → Add Server.
Example prompt (in Claude Code): Use markitup_generate to process the screenshot at https://example.com/dashboard.png —highlight the export button and use the bold_marketing template to use it for a marketing landing page.
Environment variables and local development Never commit MARKITUP_API_KEY to source control. If the key is accidentally leaked, immediately revoke it at https://markitup.app/dashboard and create a new key. Keys are stored server-side as SHA-256 hashes, and the original key is shown only once at creation time.
Error handling The server returns clear, readable errors for common situations: MARKITUP_API_KEY is invalid or missing (set or rotate the key); insufficient MarkItUp credits (top up at https://markitup.app/credits ); rate limit exceeded (wait a moment and retry).
Applicable scenarios and limitations Suitable for generating marketing visuals from screenshots, regenerating variations, outpainting, and removing backgrounds in MCP-compatible AI clients. Usage is constrained by credits and subscription status; generation, regeneration, outpainting, and background removal all consume credits (background removal is free for active Pro/Power subscribers). A valid MarkItUp account and API key are required.
