pluggedin-mcp-proxy

import { GetPromptResultSchema, PromptMessage } from "@modelcontextprotocol/sdk/types.js"; import { readFileSync } from 'fs'; import { createRequire } from 'module'; const require = createRequire(import.meta.url); const packageJson = require('../../package.json'); /** * Static prompts for the MCP proxy */ export const staticPrompts = { "mcp_discover": { name: "mcp_discover", description: "Comprehensive guide for discovering and using MCP tools in Plugged.in", arguments: [] }, "how_to_use_discovery": { name: "how_to_use_discovery", description: "Step-by-step instructions for discovering and connecting to MCP servers", arguments: [] }, "what_pluggedin_can_do_for_me": { name: "what_pluggedin_can_do_for_me", description: "Discover all the powerful features and tools Plugged.in offers to enhance your AI workflows", arguments: [] } }; /** * Get a static prompt by name */ export function getStaticPrompt(name: string): any | null { switch (name) { case "mcp_discover": return { description: staticPrompts.mcp_discover.description, messages: [{ role: "user", content: { type: "text", text: `# Plugged.in MCP Discovery Guide (v${packageJson.version}) ## 🚀 Welcome to Plugged.in MCP Proxy! I help you connect to and use multiple MCP (Model Context Protocol) servers through a single, unified interface. Think of me as your gateway to a world of specialized AI tools and capabilities. ## 🔍 Discovery Process ### Step 1: Discover Available Tools Start by running the \`pluggedin_discover_tools\` tool. This will: - Connect to your configured MCP servers - Retrieve all available tools, resources, and prompts - Present them in an organized, prefixed format ### Step 2: Understanding Tool Names All discovered tools are prefixed with their server UUID to avoid conflicts: - Format: \`{server_uuid}__{tool_name}\` - Example: \`550e8400-e29b-41d4-a716-446655440000__read_file\`, \`123e4567-e89b-12d3-a456-426614174000__create_issue\` - This ensures unique tool names even when multiple servers have tools with identical names ### Step 3: Using Discovered Tools Once discovered, simply call any tool by its prefixed name with appropriate arguments. ## 📋 Static Tools Always Available These tools are built into Plugged.in and always available: ### Discovery & Information 1. **pluggedin_discover_tools** - Discover all available MCP tools - Optional: \`server_uuid\` to discover specific server - Optional: \`force_refresh\` to bypass cache ### RAG & Knowledge Management 2. **pluggedin_ask_knowledge_base** - Query your document knowledge base - Required: \`query\` - Your search query ### Notification System 3. **pluggedin_send_notification** - Send custom notifications 4. **pluggedin_list_notifications** - List your notifications 5. **pluggedin_mark_notification_done** - Mark notifications as done 6. **pluggedin_delete_notification** - Delete notifications ### Document Management 7. **pluggedin_create_document** - Create AI-generated documents 8. **pluggedin_list_documents** - List documents with filters 9. **pluggedin_search_documents** - Search documents semantically 10. **pluggedin_get_document** - Retrieve specific documents 11. **pluggedin_update_document** - Update existing documents ## 💡 Pro Tips 1. **First Time?** Always run \`pluggedin_discover_tools\` first 2. **Performance**: Discovery results are cached for efficiency 3. **Multiple Servers**: Tools from different servers work seamlessly together 4. **Notifications**: All tool usage is logged for your reference 5. **RAG Integration**: Documents you create are automatically indexed ## 🔧 Troubleshooting - **No tools found?** Check your MCP server configuration in Plugged.in - **Tool not working?** Ensure the server is properly connected - **Need help?** The discovery tool shows server status and capabilities ## 📚 Example Workflow \`\`\` 1. Discover tools: pluggedin_discover_tools 2. Create a document: pluggedin_create_document 3. Search documents: pluggedin_search_documents 4. Use server tools: 550e8400-e29b-41d4-a716-446655440000__read_file, 123e4567-e89b-12d3-a456-426614174000__create_pr, etc. \`\`\` Ready to explore? Start with \`pluggedin_discover_tools\`!` } } as PromptMessage] }; case "how_to_use_discovery": return { description: staticPrompts.how_to_use_discovery.description, messages: [{ role: "user", content: { type: "text", text: `# How to Use MCP Discovery in Plugged.in ## Quick Start ### 1. Run Discovery \`\`\` Tool: pluggedin_discover_tools Arguments: {} (or {"force_refresh": true} to bypass cache) \`\`\` ### 2. Review Available Tools The discovery will show: - **Server Name & Status** - **Available Tools** with descriptions - **Custom Instructions** (if any) - **Resources & Prompts** (if available) ### 3. Use Discovered Tools Call any tool using its prefixed name: \`\`\` Tool: servername_toolname Arguments: { ...tool specific arguments... } \`\`\` ## Understanding the Output ### Tool Format Each tool is displayed as: \`\`\` - **servername_toolname**: Tool description \`\`\` ### Server Organization \`\`\` ## Server Name (UUID) ### Tools (count): - tool list... ### Instructions (count): - instruction list... \`\`\` ## Advanced Usage ### 1. Discover Specific Server \`\`\` Tool: pluggedin_discover_tools Arguments: {"server_uuid": "specific-server-uuid"} \`\`\` ### 2. Force Refresh (Bypass Cache) \`\`\` Tool: pluggedin_discover_tools Arguments: {"force_refresh": true} \`\`\` ### 3. Combine with Document Creation After discovery, you can create documents about discovered tools: \`\`\` Tool: pluggedin_create_document Arguments: { "title": "Available MCP Tools Reference", "content": "...discovered tools documentation...", "format": "md", "metadata": { "model": {"name": "assistant", "provider": "anthropic"}, "visibility": "private" } } \`\`\` ## Common Patterns ### Pattern 1: Discover → Use 1. \`pluggedin_discover_tools\` - Find available tools 2. \`550e8400-e29b-41d4-a716-446655440000__read_file\` - Use a discovered tool ### Pattern 2: Discover → Document → Search 1. \`pluggedin_discover_tools\` - Find tools 2. \`pluggedin_create_document\` - Document findings 3. \`pluggedin_search_documents\` - Search later ### Pattern 3: Notification Workflow 1. \`pluggedin_discover_tools\` - Initial discovery 2. Use various tools... 3. \`pluggedin_list_notifications\` - Check activity log ## Tips & Best Practices 1. **Cache Management**: Discovery results are cached for performance. Use \`force_refresh\` when servers are updated. 2. **UUID Prefixing**: Tools are prefixed with server UUIDs to avoid conflicts. A "read" tool from server "550e8400-e29b-41d4-a716-446655440000" becomes \`550e8400-e29b-41d4-a716-446655440000__read\`. 3. **Error Handling**: If a tool fails, check: - Is the server still connected? - Are the arguments correct? - Run discovery again to refresh 4. **Documentation**: Create documents about useful tool combinations for future reference. 5. **Activity Tracking**: All tool usage is logged. Check notifications to review your activity. Need more help? Just ask!` } } as PromptMessage] }; case "what_pluggedin_can_do_for_me": return { description: staticPrompts.what_pluggedin_can_do_for_me.description, messages: [{ role: "user", content: { type: "text", text: `# What Plugged.in Can Do For Me 🚀 Plugged.in transforms how you work with AI by providing a unified platform for managing MCP servers, sharing knowledge, and collaborating with AI models. Here's everything I can help you with: ## 🔧 Built-in Tools (12 Powerful Features) ### 📡 **Discovery & Management** #### 1. **pluggedin_discover_tools** - **What it does**: Discover all available tools from your connected MCP servers - **Parameters**: - \`server_uuid\` (optional): Target specific server or discover all - \`force_refresh\` (optional): Force refresh to get latest tools - **Why use it**: See all your available AI capabilities in one place ### 🆘 **Setup & Help** #### 2. **pluggedin_setup** - **What it does**: Get help setting up and using Plugged.in (no API key required) - **Parameters**: - \`topic\` (optional): getting_started, api_key, configuration, or troubleshooting - **Why use it**: Learn how to configure and use Plugged.in effectively ### 📚 **Document & Knowledge Management** #### 3. **pluggedin_ask_knowledge_base** - **What it does**: Search through your entire knowledge base using AI - **Parameters**: - \`query\` (required): Your search question (1-1000 characters) - **Why use it**: Find information across all your documents instantly #### 4. **pluggedin_create_document** - **What it does**: Save AI-generated content directly to your library - **Parameters**: - \`title\` (required): Document title - \`content\` (required): The content to save - \`format\` (optional): md, txt, json, or html - \`tags\` (optional): Tags for organization - \`metadata\` (required): AI model information and visibility settings - **Why use it**: Build a persistent knowledge base from AI conversations #### 5. **pluggedin_list_documents** - **What it does**: Browse your document library with smart filters - **Parameters**: - \`filters\` (optional): Filter by source, model, date, tags, visibility - \`sort\` (optional): Sort by date, title, or size - \`limit\` & \`offset\` (optional): Pagination controls - **Why use it**: Quickly find documents created by specific AI models or on certain topics #### 6. **pluggedin_search_documents** - **What it does**: Perform semantic search across all your documents - **Parameters**: - \`query\` (required): Search query - \`filters\` (optional): Narrow results by model, date, tags - \`limit\` (optional): Number of results - **Why use it**: Find relevant content even when you don't remember exact keywords #### 7. **pluggedin_get_document** - **What it does**: Retrieve a specific document with full details - **Parameters**: - \`documentId\` (required): Document UUID - \`includeContent\` (optional): Get the full content - \`includeVersions\` (optional): See version history - **Why use it**: Access complete document information including AI attribution #### 8. **pluggedin_update_document** - **What it does**: Update or append to existing AI-generated documents - **Parameters**: - \`documentId\` (required): Document UUID - \`operation\` (required): replace, append, or prepend - \`content\` (required): New content - \`metadata\` (optional): Update tags and AI model info - **Why use it**: Evolve documents over time with contributions from multiple AI models ### 🔔 **Notification System** #### 9. **pluggedin_send_notification** - **What it does**: Send custom notifications with optional email delivery - **Parameters**: - \`message\` (required): Notification content - \`title\` (optional): Custom title - \`severity\` (optional): INFO, SUCCESS, WARNING, or ALERT - \`email\` (optional): Also send via email - **Why use it**: Stay informed about important events and completions #### 10. **pluggedin_list_notifications** - **What it does**: View your notification history - **Parameters**: - \`unreadOnly\` (optional): Filter unread only - \`limit\` (optional): Number to retrieve (1-100) - \`severity\` (optional): Filter by severity level - **Why use it**: Never miss important updates from your AI workflows #### 11. **pluggedin_mark_notification_done** - **What it does**: Mark notifications as done - **Parameters**: - \`notificationId\` (required): Notification ID - **Why use it**: Keep your notification center organized #### 12. **pluggedin_delete_notification** - **What it does**: Remove notifications - **Parameters**: - \`notificationId\` (required): Notification ID - **Why use it**: Clean up processed notifications ## 🌟 Key Platform Features ### 🤖 **AI Document Exchange** - **Cross-Model Collaboration**: Different AI models can build on each other's work - **Version Control**: Track changes and contributions from multiple AI models - **Attribution System**: Know which AI created or updated each document - **Visibility Controls**: Keep documents private, share with workspace, or make public ### 🔐 **Security & Organization** - **Profile-based Isolation**: Each workspace has its own isolated environment - **API Key Authentication**: Secure access to all features - **Activity Logging**: Complete audit trail of all operations - **Rate Limiting**: Protection against abuse ### 🚀 **Getting Started** 1. **Set Up**: Configure \`PLUGGEDIN_API_KEY\` and \`PLUGGEDIN_API_BASE_URL\` 2. **Discover**: Use \`pluggedin_discover_tools\` to see all available capabilities 3. **Create**: Start building your AI knowledge base with \`pluggedin_create_document\` 4. **Search**: Find information instantly with \`pluggedin_ask_knowledge_base\` or \`pluggedin_search_documents\` 5. **Collaborate**: Let multiple AI models contribute to your documents ## 💡 **Common Use Cases** ### Building a Knowledge Base - Save important AI conversations as documents - Tag and organize by topic - Search across all saved knowledge - Update documents as you learn more ### AI Collaboration Workflows - One AI creates initial documentation - Another AI reviews and appends improvements - Track all contributions with full attribution - Build comprehensive resources over time ### Research & Analysis - Store research findings from different AI models - Search across all collected information - Version control for evolving insights - Share findings with your team ## 🎯 **Pro Tips** 1. **Use Tags**: Organize documents with meaningful tags for easy filtering 2. **Leverage Search**: Both RAG and semantic search help find information differently 3. **Version Everything**: AI-generated documents automatically track versions 4. **Set Visibility**: Choose who can see your documents (private, workspace, public) 5. **Check Notifications**: Stay updated on all activities in your workspace Plugged.in transforms isolated AI conversations into a persistent, searchable, and collaborative knowledge ecosystem. Start exploring what it can do for you today!` } } as PromptMessage] }; default: return null; } }