MCP Proof of Concept

Prerequisites

• UV
• Docker

Installing UV

UV is a package/project manager for python. It's great.

# RBAC-matcher: Org Chart AI via MCP

## Overview

This project enables AI-powered exploration and visualization of org structures
from Excel files using a modular MCP (Multi-Channel Protocol) server and a
command-line client. It includes fuzzy Excel file search, org chart generation,
and OpenAI LLM-driven interactions.

---

## Quick Start for Developers

### 1. Prerequisites

- [**Python**](https://www.python.org/downloads/) (3.9+)
- [**UV**](https://github.com/astral-sh/uv) (Fast Python package/project manager)
- [**Docker Desktop**](https://docs.docker.com/desktop/) (for containerized file access and persistence)
- A valid OpenAI or Azure OpenAI API key

### 2. Setup

1. **Clone the repository:**
   ```bash
   git clone <this-repo-url>
   cd RBAC-matcher

2. Install UV:

powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"

Or see [uv install docs](https://github.com/astral-sh/uv#installation).

3. **Install Docker Desktop:**
- [Get Docker](https://docs.docker.com/desktop/setup/install/).

4. **Configure Environment:**
- Copy `.env.example` to `.env` and fill out all fields:
   - Both API keys should be the same if using only OpenAI.
   - Set endpoint URLs and versions as appropriate for your OpenAI deployment.

5. **(Optional) Set Your `sec_edgar_user_agent`:**
- If required by your dependencies or environments, edit it accordingly (see source code).

---

### 3. Installation

```bash
uv sync

Docker Build and Run:

docker build -t rbac_matcher .
docker container create --name rbac_matcher_container -p 3000:3000 rbac_matcher
docker container start rbac_matcher_container

(You may mount your /data volume as needed for Excel files.)

---

4. Running the Server

Starts on port 3000 by default:

uv run server.py
# Or use Docker as above

---

5. Running the Client

uv run client.py

• Enter server port number (3000 by default).
• At the > prompt, enter your natural language command (e.g., "Show the org chart for Workday_Data.xlsx"). Type exit to quit.

---

Features

• Fuzzy Excel File Search: You don’t need to remember or type the exact filename; a close match to any allowed Excel file in /data, /app, or current directory works.
• Automated Org Chart Visualization: Interactive HTML org charts saved alongside the Excel file.
• Natural Language AI Agent: Driven by persona and LLM backend; seamless query interface.

---

Directory Layout

RBAC-matcher/
├── server.py        # MCP server (org chart tool with fuzzy search)
├── client.py        # CLI client agent
├── .env.example     # Environment variable template
├── personas/        # Persona instructions for AI agent
├── requirements.txt # Python dependencies (if exists)
├── Dockerfile
├── /data/           # (Optional) Excel files directory

---

Troubleshooting

• Excel file not found?
  • Make sure your file is inside /data (if using Docker), /app, or your current working directory.
  • Typos are OK! Fuzzy search will pick the closest file.
• Dependency errors?
  • Run uv sync after pulling any updates.
• Org chart missing?
  • Output HTML will be created (e.g., <YOUR_FILE>_org_chart.html) in the data directory.
• API errors?
  • Double check your .env values and network access.

---

Developer Notes

• See top-level docstrings in server.py and client.py for code-level orientation.
• For MCP/AI/agent extension, see the agents directory and associated classes.

System Design

---

Contributing

Issues and PRs welcome! Please ensure you adhere to typical Python formatting and include test/demo files if you update features.

### Installing Docker

https://docs.docker.com/desktop/setup/install/windows-install/

## Setting up Server

First, change the `sec_edgar_user_agent` on line 19.

Next, change copy the `.env.example` to `.env` and fill out the fields (both API keys should be the same)

Run these commands:
1. `uv sync`
2. `docker build -t mcp_docker_test .`
3. `docker container create --name mcp_container -p 3000:3000 mcp_docker_test` 
4. `docker container start mcp_container`

If you open the Docker Desktop App, you should see the container running. 

## Running client

1. `uv run client.py`
2. Then enter the port you specified
3. When you see the `>` prompt, you should be all set to prompt!
4. To exit without issue, just type `exit`