toyMCP To-Do List Server

ToyMCP is a lightweight, production‑ready example of the Model Context Protocol (MCP) in action. It demonstrates how an AI assistant can be wired to a real‑world data store—here, a PostgreSQL database that holds simple to‑do items—via a JSON‑RPC 2.0 interface. The server exposes CRUD operations for to‑do entries, allowing an assistant to create, read, update, or delete tasks based purely on natural‑language prompts. This bridge removes the need for developers to hand‑craft HTTP calls or manage authentication tokens manually, letting the assistant orchestrate data flows as if it were a first‑class citizen of the application.

For developers, the value lies in the seamless integration with existing AI workflows. The MCP server presents a clean, typed contract that an assistant can introspect: each method’s parameters and return shape are documented in the JSON‑RPC schema, and Swagger UI provides a human‑readable view. Once authenticated via JWT, an assistant can invoke , , or without any intermediate translation layer. This reduces boilerplate, speeds up prototyping, and ensures that the assistant’s actions are auditable against a persistent backend.

Key capabilities include:

• JSON‑RPC over HTTP – A lightweight, request/response protocol that is easy to serialize and debug.
• JWT authentication – Securely ties assistant actions to a user context, enabling fine‑grained access control.
• PostgreSQL persistence – Guarantees durability and relational integrity for to‑do items.
• Swagger UI documentation – Gives developers instant visibility into the API contract and live testing tools.
• Docker‑based deployment – The database runs in a container, simplifying local setup and CI pipelines.

Typical use cases are plentiful: an assistant can act as a personal productivity coach, automatically adding tasks from email or calendar events; it can serve as the backend for a voice‑controlled home assistant that manages shopping lists; or it can power an internal ticketing system where employees ask the AI to log issues in a shared repository. In each scenario, the MCP server abstracts away database intricacies, letting developers focus on intent modeling and user experience.

What sets ToyMCP apart is its educational clarity. By coupling a minimal Express server with well‑documented JSON‑RPC endpoints, it offers a hands‑on reference for building more complex MCP servers—whether you need advanced filtering, batch operations, or real‑time notifications. The modular design also encourages extending the agent framework to support additional tools (e.g., calendar APIs or file storage) without changing the underlying protocol.