In this article
September 3, 2025
September 3, 2025

From PRD to Prototype in One Prompt: How Notion's MCP Server Transforms Product Development

Marissa Felix and Kenneth Sinder from Notion's API team demonstrate how their MCP server bridges the gap between documentation and implementation, enabling developers to go from product requirements to working code with a single prompt.

This post is part of our ongoing MCP Night 2.0 series, highlighting the impressive demonstrations from the event. Check out our full event recap and other demo spotlights to see how the Model Context Protocol ecosystem is evolving.

When Marissa Felix and Kenneth Sinder from Notion took the stage at MCP Night 2.0, they opened with a simple question:

"How many of you here use Notion today?" Nearly every hand in the room went up.

What followed was a demonstration that showed why Notion's MCP server might be one of the most practical integrations for everyday developers.

The Demo That Made Everyone Take Notice

Marissa began with a compelling scenario that resonates with anyone who's worked in product development. She had written a PRD (Product Requirements Document) for a rainbow memory game in Notion. Rather than manually translating requirements into code, she simply jumped into Cursor and asked: "Hey, can you implement this PRD for me?"

What happened next was remarkable. The system automatically pulled the complete PRD from Notion, understood the requirements, and generated a fully functional memory game. In one prompt, using Notion's MCP server, they had a working prototype.

But the real power became clear when Marissa explained the bidirectional nature of this integration. "You can imagine that we could extend this same exact use case to say I already have some code already implemented," she noted. "I can also prompt Cursor to say, hey, did I miss anything in this PRD or even update the PRD itself to reflect the actual product functionality."

This creates a living connection between documentation and implementation—something most development teams struggle to maintain.

The Evolution of Notion's MCP Implementation

Kenneth provided valuable insight into Notion's journey with MCP, revealing how they evolved from a basic integration to a sophisticated, production-ready system.

Version 1: The Self-Hosted Beginning

Notion's initial MCP implementation started as a downloadable, open-source package. Users had to host it themselves, create API bearer tokens, and provision them with Notion to access a subset of public APIs. While functional, this approach had significant limitations.

"We heard from the community that with this solution, you kind of hit context token limits pretty quickly because of the back and forth tool calls that are made when you fetch a Notion page incrementally layer by layer," Kenneth explained.

The technical challenge was rooted in Notion's complex schema and the way pages are structured with nested block children arrays in JSON. This created a chatty API interaction that consumed context tokens rapidly.

Version 2: The Remote Revolution

Recognizing these limitations, Notion built their remote MCP server from the ground up, designed specifically for "single player human in the loop AI agent interaction experience."

This new approach brought several major improvements:

Simplified Installation: Instead of self-hosting, users can now OAuth into their workspace through a common Notion MCP public integration. The installation process became significantly easier.

Enhanced Markdown Support: The new system can create, update, and fetch pages using an enhanced Markdown syntax that goes beyond CommonMark. This adds the expressiveness users expect from Notion—colors, formatting, and all the rich content types that make Notion documents powerful.

User-Based Permissions: Unlike the previous bot permission model, the new system uses user-based permissions. It can access all the pages and databases that you can access, and actions like creating pages are attributed to you rather than a bot.

Modern Infrastructure: Kenneth revealed their technical stack, built on Anthropic packages and Hono server, hosted on Cloudflare Workers with durable objects and KV store for managing token refresh. This infrastructure choice helped them "move quickly and focus on our business logic."

The One-Click Installation Success

One of Notion's proudest achievements has been the adoption across major MCP client directories. Kenneth highlighted their presence across the ecosystem: "One of the highlights of our launch moment last month was having a bunch of the prominent MCP client directories feature us as an MCP server that you can do this one-click installation with."

The list is impressive: ChatGPT, Claude, Cursor, and VS Code all feature Notion as a one-click MCP installation. This broad compatibility demonstrates both the quality of their implementation and the demand for Notion integration across different development environments.

Real-World Applications

The Notion MCP server enables several compelling use cases:

Documentation-Driven Development: Pull PRDs, technical specifications, and design documents directly into your development environment, ensuring code matches documented requirements.

Living Documentation: Update Notion documents based on implemented code, creating a feedback loop that keeps documentation current with actual functionality.

Cross-Workspace Search: Leverage Notion's enterprise search capabilities to find relevant information across your entire workspace and connected tools.

Project Coordination: Access and update project databases, task lists, and team documentation without leaving your development environment.

Challenges and Future Directions

Kenneth was candid about the limitations they've encountered in the current MCP ecosystem. Some MCP clients like N8N don't work well with stateful servers, and they've hit tool limits with certain clients after adding just a few dozen tools from different providers.

"You can imagine that if some of these MCP servers are using very granular modular small tools, you can kind of hit that limit quickly," he noted. They've also encountered issues with complex schemas causing exceptions in some clients.

The Roadmap Ahead

Notion's team is actively working on several fronts to improve both their MCP server and the broader ecosystem:

Ecosystem Improvements: They want to work with other steering committee members to improve server discovery protocols and installation simplicity. Kenneth painted a picture of being able to go from a Notion page URL to automatically discovering and installing the relevant MCP server.

Enhanced Security and Reliability: Focus on making recommendations for evaluations, security practices, and reliability standards across the MCP ecosystem.

Expanded Functionality: Adding more tools and better support for multiplayer scenarios, along with exploring what business models might look like for workflows involving headless agents.

The Technical Architecture

For developers interested in the implementation details, Notion's architecture choices offer valuable lessons:

  • Cloudflare Workers for hosting provides global distribution and serverless scaling
  • Durable Objects with KV store handle stateful operations like token management
  • Hono server framework provides a modern, lightweight HTTP server foundation
  • Standard Anthropic packages ensure compatibility with the broader MCP ecosystem

This stack enables Notion to focus on their core business logic while leveraging proven infrastructure for the operational challenges of running a production MCP server.

Why This Matters for Product Development

Notion's MCP implementation represents something significant for how product teams work. The traditional handoff between product requirements and implementation often introduces friction, miscommunication, and drift between what's documented and what's built.

By creating a direct, bidirectional connection between Notion documents and development environments, they're enabling a new kind of workflow where documentation and code can stay synchronized through the development process.

The demo of going from PRD to working prototype in one prompt isn't just impressive—it's a glimpse of how AI-enabled development might evolve. When documentation becomes directly actionable and implementation can automatically update documentation, the traditional barriers between product and engineering start to dissolve.

Getting Started with Notion MCP

For teams interested in trying Notion's MCP server, installation is now straightforward across major development environments. The one-click installation process works with Cursor, VS Code, Claude, and other popular MCP clients.

The integration respects your existing Notion permissions and workflows, so you can start experimenting without disrupting your current processes. Whether you're pulling project requirements into your IDE or updating documentation based on implemented features, Notion MCP offers a practical entry point into AI-augmented development workflows.

This post is part of our MCP Night 2.0 demo recap series. Watch Marissa and Kenneth's full presentation on YouTube to see the PRD-to-prototype demo in action.

This site uses cookies to improve your experience. Please accept the use of cookies on this site. You can review our cookie policy here and our privacy policy here. If you choose to refuse, functionality of this site will be limited.