Stytch Consumer Todo List MCP Server

The Workers + Stytch TODO App MCP Server demonstrates how a conventional full‑stack application can be extended to serve AI assistants through the Model Context Protocol. At its core, the server combines a static React front‑end, a lightweight REST API backed by Cloudflare KV, and an MCP server powered by Durable Objects. This architecture lets Claude or other AI agents query and manipulate a user’s to‑do list while preserving the same authentication flow that users already trust.

The primary problem this MCP solves is the friction between human‑centric web apps and AI workflows. Traditional applications expose REST endpoints, but AI assistants require a conversational interface that can maintain context across interactions. By exposing the same data through MCP, an assistant can ask a user “What are my tasks for tomorrow?” and receive structured, authenticated responses without the need to build a separate backend. Developers can therefore unlock AI‑driven productivity features—such as auto‑generated summaries, intelligent task prioritization, or voice‑activated updates—using the existing application logic.

Key capabilities of this server include:

• User authentication via Stytch Consumer: Secure, password‑less login with email magic links or SMS OTPs, ensuring that only authorized users can access their to‑do data.
• Stateful interactions with Durable Objects: Each user gets a dedicated object that preserves conversation history and task state across requests, enabling the AI to reason about past actions.
• Fast data access with Workers KV: CRUD operations on tasks are served from key‑value storage, providing low latency and horizontal scalability.
• Model Context Protocol surface: The endpoint exposes resources, prompts, and tools that Claude can invoke, turning ordinary API calls into rich conversational actions.

Real‑world scenarios abound: a customer support bot could pull a user’s pending tickets, an email assistant might suggest follow‑up tasks, or a project manager could let team members query sprint backlogs through natural language. Because the MCP layer sits on top of existing infrastructure, developers can incrementally adopt AI features without refactoring their entire stack.

What sets this demo apart is its tight integration of Cloudflare Workers, Durable Objects, and Stytch’s consumer‑focused authentication. The result is a production‑ready, serverless solution that scales with user demand while keeping the developer experience straightforward. By exposing familiar REST patterns through MCP, teams can rapidly prototype AI extensions that feel native to both users and assistants.