Lunch Money MCP Server

Lunch Money MCP Server

LLM agent-driven development of a Model Context Protocol server for Lunch Money API access through systematic agent execution with rigorous validation processes.

🎯 Repository Goal

This repository builds a Model Context Protocol (MCP) server that provides seamless access to Lunch Money financial data via standard IO (stdio) transport.

Goal 1: Enable AI assistants to interact directly with Lunch Money's API through standardized MCP tools using stdio (not remote), allowing users to:

• Query transaction data with flexible filtering
• Access spending categories and budget information
• Retrieve transaction tags and organizational data
• Perform financial analysis through natural language

🚧 Work in Progress

This project is actively under development using a systematic agent execution approach. Every line of code, configuration, and documentation is implemented through LLM agents following structured workflows.

🤖 LLM Agent-Driven Development

This repository showcases a novel development methodology where:

• LLM agents execute all coding tasks following predefined rules and validation checkpoints
• No manual coding - agents handle implementation, testing, and documentation
• Systematic validation ensures quality through mandatory human approval at each step
• Structured task management breaks complex features into validated sub-tasks

Agent Execution Framework

Significant engineering effort has been invested in creating comprehensive rules and processes that enable:

• Self-executing agents that can autonomously implement features
• Clear validation marks with mandatory human approval between sub-tasks
• Quality assurance through structured TDD and testing requirements
• Systematic progression from PRD → TDD → Tasks → Implementation

The agent execution rules in /rules/ define:

• Task breakdown and dependency management
• Validation checkpoints and quality gates
• Branch management and PR generation
• Error handling and feedback loops

📁 Project Structure

🔧 Technology Stack

• Runtime: Bun (fast TypeScript execution)
• Framework: Model Context Protocol SDK
• Validation: Zod schemas
• HTTP Client: Axios
• Testing: Built-in bun test runner

This README will be updated as the project progresses through agent-driven development milestones.