Optimizing Notion Workspaces with NoteRunway and the Model Context Protocol

NoteRunway: Because Your Notion Workspace Deserves an Elite Crew

NoteRunway is an AI-powered flight control tower for Notion workspaces built for the Notion MCP Challenge. The system utilizes seven specialized tools to scan for orphaned pages, dead links, and sensitive API keys across complex database structures.

Why This Matters

Power users frequently suffer from ‘workspace rot,’ where manual management of hundreds of nested pages becomes impossible, leading to orphaned content and security vulnerabilities. NoteRunway addresses this by implementing a hybrid architecture that uses the direct Notion SDK for high-speed bulk reads and the Model Context Protocol (MCP) as a sandboxed execution layer for AI-driven modifications. This technical reality ensures that while an AI can reason about workspace architecture, destructive actions are governed by a human-in-the-loop protocol, preventing the data loss common in fully automated bots.

Key Insights

• Semantic duplicate detection uses LLMs to group pages with overlapping content even if titles differ, assigning similarity scores for manual review.
• The Sensitive Data Finder employs a two-phase security scanner using 13 regex patterns for keys (AWS, Stripe, GitHub) followed by an AI deep scan for natural language secrets.
• The system architecture routes agentic chat through Vercel AI SDK and MCP Client to the notion-mcp-server via stdio (JSON-RPC 2.0).
• Workspace visualization is handled by React Flow 11, creating interactive force-directed graphs where nodes are colored by depth and orphaned pages glow red.
• NoteRunway implements a ‘Bring Your Own Key’ (BYOK) privacy model where AI keys are stored only in the browser’s localStorage and never persisted server-side.

Working Examples

The execution architecture for the Semantic Ask agentic AI loop.

Browser → Next.js API (SSE) → Vercel AI SDK → MCP Client → notion-mcp-server (stdio) → Notion API

Practical Applications

• Automated Workspace Auditing: NoteRunway identifies stale pages untouched for 90+ days. Pitfall: Deleting content without recovery paths is avoided by creating ‘Audit Stubs’ in a dedicated archive folder before soft-archiving to Notion’s trash.
• Security Hardening: Scanning meeting notes for accidentally pasted OpenAI or Stripe keys. Pitfall: Standard regex scanners miss plain-text passwords, which NoteRunway catches via LLM-based natural language analysis.
• Dependency Management: Visualizing parent-child relationships and broken @mentions in large documentation sets to find disconnected nodes that should be re-linked or archived.

References:

• https://dev.to/georgekobaidze/noterunway-because-your-notion-workspace-deserves-an-elite-crew-53bk
• https://noterunway.pilotronica.com