🧩 NYT Wordle AI Solver

A LangGraph-powered AI agent built to solve the daily New York Times Wordle puzzle using a structured workflow and MCP-enabled publishing.

---

✨ Features

• 🧠 Graph-Based Intelligence: Uses a LangGraph workflow to manage guessing, validation, and publishing.
• 🛠️ MCP Integration: Publishes results through external MCP tools (Slack, WhatsApp, NYT).
• 💬 Slack Publishing: Sends the Wordle result grid to a configured Slack channel.
• 📱 WhatsApp Integration: Automated WhatsApp messaging using pywhatkit.
• 📝 NYT Commenting: Generates and posts friendly comments for the NYT community.
• 🎨 Rich Terminal UI: Provides live feedback and polished terminal output.
• 🧱 Modular Codebase: Clear separation between game logic, prompts, solver nodes, and publishing.

---

🚀 Project Structure

• src/app.py — main application entrypoint
• src/core/game.py — Wordle game rules, feedback, and grid rendering
• src/core/graph_builder.py — LangGraph workflow construction and node transitions
• src/agents/prompts.py — LLM prompt templates and chain configuration
• src/agents/solver.py — solver node implementations for guessing and validation
• src/agents/publisher.py — MCP publisher node
• src/mcp/social-media-mcp.py — MCP server hosting Slack, WhatsApp, and NYT tools
• src/mcp/whatsapp.py — WhatsApp messaging utility

---

🏗️ Architecture

The solver runs as a three-node state machine:

graph TD;
    __start__([__start__])
    GUESS(GUESS)
    VALIDATE(VALIDATE)
    PUBLISH(PUBLISH)
    __end__([__end__])

    __start__ --> GUESS;
    GUESS --> VALIDATE;
    VALIDATE -.-> GUESS;
    VALIDATE -.-> PUBLISH;
    PUBLISH --> __end__;

Loading

---

Getting Started

Prerequisites

• uv
• A compatible LLM backend configured via MODEL_NAME
• A local or remote MCP endpoint

Installation

1. Clone the repository:

   git clone https://github.com/Navin3d/NYT-Wordle-Solver.git
   cd NYT-Wordle-Solver

2. Install dependencies:

   uv sync

3. Configure environment variables by creating src/.env:

   SLACK_BOT_TOKEN=xoxb-111111-22221222-jhghg
   SLACK_CHANNEL_ID=#general
   WHATSAPP_CONTACTS_TO_SEND=+919442807217
   MODEL_NAME=gemma4:latest

🛠️ Running the MCP Server

Before running the solver, start the Social Media MCP server in a separate terminal:

uv run python src/mcp/social-media-mcp.py

The current publisher module is configured to use an MCP service at http://localhost:8010/mcp.

---

Usage

Run the solver from the project root:

uv run python src/app.py

---

Tech Stack

• LangGraph / LangChain
• Python 3.13
• Rich
• uv
• FastMCP

---

References

• NYT Wordle JSON endpoint
• Fast MCP Quickstart
• Mermaid.js