Give your AI assistant superpowers for Zotero plugin development. 28 tools for screenshots, DOM inspection, JavaScript execution, build integration, and debugging via Model Context Protocol.
README
Give your AI assistant superpowers for Zotero plugin development
Architecture · Getting Started · Available Tools
<img src="docs/images/demo.png" alt="MCP Server Zotero Dev in action" width="800"> </div>A Model Context Protocol (MCP) server that enables AI assistants like Claude, Cursor, and Windsurf to build, test, and debug Zotero 7, 8, and 9 plugins. Screenshots, DOM state, debug logs, and JavaScript execution give the AI rich context to understand what's happening—and tools to help you fix it.
| Category | Capabilities | |----------|-------------| | 🎯 UI Inspection | Screenshots, DOM tree, element finding, computed styles | | 💻 JS Execution | Run code in Zotero context, inspect APIs, test snippets | | 🔧 Build Tools | Scaffold integration for build, serve, hot reload | | 📋 Logs & Errors | Stream debug output, error console, watch for issues | | 🗃️ Database | Read-only access to zotero.sqlite for debugging | | 🔌 Plugin Management | Install, reload, list plugins |
Use install-mcp to add the server to your AI assistant:
npx -y install-mcp @introfini/mcp-server-zotero-dev --client claude-code
Supported clients: claude-code, cursor, windsurf, vscode, cline, roo-cline, claude, zed, goose, warp, `co