AI File Assistant

A document intelligence system that enables AI agents and users to read, search, summarize, and answer questions over local PDF, DOCX, XLSX, and TXT documents.

The project exposes the same core engine through:

• MCP Server for Claude Desktop and other MCP-compatible clients
• Flask Web Application for browser-based interaction

---

Features

• Read PDF, DOCX, XLSX and TXT documents
• Search keywords across multiple documents
• AI-powered document summarization
• Retrieval-Augmented Document Q&A
• Upload and manage files
• Delete individual or all documents
• Secure file access with path validation
• Claude Desktop MCP integration

---

Tech Stack

Category	Technologies
Backend	Python, Flask
MCP	Model Context Protocol (MCP)
AI	Groq Llama / Ollama
Document Parsing	PyMuPDF, python-docx, pandas, openpyxl
Frontend	HTML, CSS, JavaScript
Testing	Pytest

---

Architecture

                 Claude Desktop
                       │
                       ▼
                 MCP Server
                       │
        ┌──────────────┴──────────────┐
        │                             │
   Shared Core Engine            Flask Web App
        │
 ├── File Readers
 ├── Search Engine
 ├── Document Processor
 └── LLM Client

📸 Screenshots

🖥️ User Interface

Modern web interface for uploading, browsing, and managing supported documents.

<img width="1600" height="884" alt="image" src="https://github.com/user-attachments/assets/1b59456b-2224-430a-8f61-52e47a47d1e7" />

---

🔗 MCP Server Connected

Claude Desktop successfully connected to the local MCP File Assistant server.

<img width="1600" height="1135" alt="image" src="https://github.com/user-attachments/assets/181a46ee-2609-486e-8e10-c78f686f3a01" />

---

📄 Read Document

Read PDF, DOCX, XLSX, and TXT documents directly through Claude Desktop using MCP.

<img width="1600" height="935" alt="image" src="https://github.com/user-attachments/assets/6c638b40-4417-4385-989f-2d1485b07ad3" /> <img width="1600" height="936" alt="image" src="https://github.com/user-attachments/assets/adf11f2e-5bba-4b0b-b0f6-3e75a5458552" />

---

📝 AI Document Summarization

Generate concise AI-powered summaries of lengthy documents using Groq Llama.

<img width="1600" height="920" alt="image" src="https://github.com/user-attachments/assets/a973326b-e272-4d58-b116-c769aeb66af4" />

---

🔍 Keyword Search

Search keywords across every indexed document with ranked search results.

<img width="1600" height="940" alt="image" src="https://github.com/user-attachments/assets/a6b0323c-4352-46af-95ca-24fd5fa474c6" />

---

❓ Document Question Answering

Ask natural language questions grounded in document contents using Retrieval-Augmented Generation (RAG).

<img width="1600" height="591" alt="image" src="https://github.com/user-attachments/assets/37fb0755-a70c-4785-b519-5e5c1cc4f8d7" />

---

🗑️ File Management

Delete individual files or remove all uploaded documents directly from the application.

<img width="1600" height="897" alt="image" src="https://github.com/user-attachments/assets/91548827-f15e-4142-acc0-2ce10f99c229" />

---

Claude Desktop MCP Integration

Claude Desktop successfully invokes custom MCP tools including Read, Search, Summarize, and Ask Document.

<img width="1600" height="941" alt="image" src="https://github.com/user-attachments/assets/07e29083-a482-4d10-bc97-a623a285a717" />

---

📁 Project Structure

mcp-file-assistant/
│
├── app.py
├── server.py
├── config.py
├── file_readers.py
├── search_engine.py
├── document_processor.py
├── llm_client.py
├── templates/
├── static/
├── tests/
├── sample_documents/
├── requirements.txt
└── README.md

---

🚀 Installation

git clone <repository-url>

cd mcp-file-assistant

pip install -r requirements.txt

python app.py

Open:

http://127.0.0.1:5000

To start the MCP server:

python server.py

Configure Claude Desktop with the provided MCP configuration and restart the application.

---

🔧 MCP Tools

• Read Document
• Search Documents
• Summarize Document
• Ask Document
• List Files
• Delete File
• Delete All Files

---