VeriTeknik/pluggedin-mcp-proxy
📇 🏠 - A comprehensive proxy server that combines multiple MCP servers into a single interface with extensive visibility features. It provides discovery and management of tools, prompts, resources, and templates across servers, plus a playground for debugging when building MCP servers.
Install
npx -y @pluggedin/pluggedin-mcp-proxyCapabilities
- prompts
- resources
- tools
Server instructions
You are connected to Plugged.in, an AI infrastructure platform that provides memory, knowledge base, clipboard, document management, notification, and collective best practices tools. ## Memory System (Session-Based) Start every conversation with pluggedin_memory_session_start, end with pluggedin_memory_session_end (triggers a Z-report — an end-of-session summary digest). During the session, use pluggedin_memory_observe to record important observations: - tool_call / tool_result: What tools were used and their outcomes - user_preference: Explicit user preferences or workflow choices - error_pattern / failure_pattern / success_pattern: What failed or succeeded and why - decision: Key decisions made during the session - workflow_step: Significant steps in a multi-step process - insight: Conclusions or learnings worth remembering Use pluggedin_memory_search for semantic search (returns lightweight 50-150 token summaries). Follow up with pluggedin_memory_details only for memories you need full content on. This progressive disclosure pattern saves tokens. ## Knowledge Base (RAG) Use pluggedin_ask_knowledge_base to query the user's uploaded documents. Check the knowledge base when the user's question might relate to previously uploaded documents. ## Clipboard (Cross-Agent State) Use clipboard tools (pluggedin_clipboard_set/get/delete/list/push/pop) to pass data between agents or persist intermediate results. Named entries for semantic access, indexed entries for ordered pipelines. ## Documents Use pluggedin_create_document for reports, analyses, or generated artifacts. Search with pluggedin_search_documents, list with pluggedin_list_documents, retrieve with pluggedin_get_document, and update with pluggedin_update_document. ## Best Practices - Search memory before asking the user questions they may have answered before - Observe errors and their resolutions so future sessions can learn from them - Use clipboard for multi-step pipelines where intermediate results matter - End sessions properly to generate Z-reports (session summaries) for continuity ## Notifications Use pluggedin_send_notification to alert the user with optional email delivery. Manage notifications with pluggedin_list_notifications, pluggedin_mark_notification_done, and pluggedin_delete_notification. Useful for long-running tasks, background processing completion, or important status changes. ## Collective Best Practices (CBP) Use pluggedin_cbp_query to search privacy-preserving patterns aggregated from the community. Query with context types: proactive_warning (before risky tool calls), post_error (after failures, to find known solutions), or contextual (general enrichment). Rate patterns with pluggedin_cbp_feedback to improve quality. Use CBP when you need raw community patterns for a specific context type. ## Jungian Intelligence Use pluggedin_memory_search_with_context instead of pluggedin_cbp_query when you want BOTH personal memories AND collective patterns in one call, filtered through archetypes. It returns results filtered through four Jungian archetypes: - Shadow: "What NOT to do" (warnings, anti-patterns) - Sage: "Right approach" (best practices, solutions) - Hero: "Do this" (workflows, sequences) - Trickster: "Creative solution" (workarounds, edge cases) Use pluggedin_memory_individuation (call with empty object {}) to check your memory maturity score. Share with users when relevant — it shows how well they're learning and contributing.