FS-MCP: Universal File Reader & Intelligent Search MCP Server

A powerful MCP (Model Context Protocol) server that provides intelligent file reading and semantic search capabilities

English | 中文

English

🚀 Features

• 🧠 Intelligent Text Detection: Automatically identifies text files without relying on file extensions

• 📄 Multi-Format Support: Handles text files and document formats (Word, Excel, PDF, etc.)

• 🔒 Security First: Restricted access to configured safe directories only

• 📏 Range Reading: Supports reading specific line ranges for large files

• 🔄 Document Conversion: Automatic conversion of documents to Markdown with caching

• 🔍 Vector Search: Semantic search powered by AI embeddings

• ⚡ High Performance: Batch processing and intelligent caching support

• 🌐 Multi-language: Supports both English and Chinese content

📋 Table of Contents

• Quick Start

• Installation

• Configuration

• MCP Tools

• Vector Search

• Supported Formats

• Security Features

• Integration

• Development

• Contributing

• License

🚀 Quick Start

1. Clone and Install

git clone https://github.com/yourusername/fs-mcp.git
cd fs-mcp

Using uv (Recommended):

uv sync

Using pip:

pip install -r requirements.txt  # If you have a requirements.txt
# OR install directly
pip install fastmcp>=2.0.0 langchain>=0.3.0 python-dotenv>=1.1.0

2. Environment Configuration

Create a .env file in the project root:

# Security Settings
SAFE_DIRECTORY=.                    # Directory restriction (required)
MAX_FILE_SIZE_MB=100                # File size limit in MB

# Encoding Settings
DEFAULT_ENCODING=utf-8

# AI Embeddings Configuration (for vector search)
OPENAI_EMBEDDINGS_API_KEY=your-api-key
OPENAI_EMBEDDINGS_BASE_URL=http://your-embedding-service/v1
EMBEDDING_MODEL_NAME=BAAI/bge-m3    # Or your preferred model
EMBEDDING_CHUNK_SIZE=1000

3. Start the Server

python main.py

The server will start on http://localhost:3002 and automatically build the vector index.

🛠️ Installation

System Requirements

• Python: 3.12 or higher

• OS: Windows, macOS, Linux

• Memory: 4GB+ recommended for vector search

• Storage: 1GB+ for caching and indexes

Dependencies

Core dependencies are managed in pyproject.toml:

• fastmcp>=2.0.0 - MCP server framework

• langchain>=0.3.0 - AI and vector search

• python-dotenv>=1.1.0 - Environment management

• Document processing libraries (pandas, openpyxl, python-docx, etc.)

⚙️ Configuration

Environment Variables

Advanced Configuration

For production deployments, consider:

• Setting up rate limiting

• Configuring log rotation

• Using external vector databases

• Setting up monitoring

🔧 MCP Tools

1. view_directory_tree

Purpose: Display directory structure in tree format

view_directory_tree(
    directory_path=".",     # Target directory
    max_depth=3,           # Maximum depth
    max_entries=300        # Maximum entries to show
)

2. read_file_content

Purpose: Read file content with line range support

read_file_content(
    file_path="example.py",  # File path
    start_line=1,           # Start line (optional)
    end_line=50             # End line (optional)
)

3. search_documents

Purpose: Intelligent semantic search across documents

search_documents(
    query="authentication logic",     # Search query
    search_type="semantic",          # semantic/filename/hybrid/extension
    file_extensions=".py,.js",       # File type filter (optional)
    max_results=10                   # Maximum results
)

4. rebuild_document_index

Purpose: Rebuild vector index for search

rebuild_document_index()  # No parameters needed

5. get_document_stats

Purpose: Get index statistics and system status

get_document_stats()  # Returns comprehensive stats

6. list_files

Purpose: List files in directory with pattern matching

list_files(
    directory_path="./src",  # Directory to list
    pattern="*.py",         # File pattern
    include_size=True       # Include file sizes
)

7. preview_file

Purpose: Quick preview of file content

preview_file(
    file_path="example.py",  # File to preview
    lines=20                # Number of lines
)

🔍 Vector Search

Capabilities

• Semantic Understanding: Search "user authentication" finds "login verification" code

• Synonym Recognition: Search "database" finds "数据库" (Chinese) content

• Multi-language Support: Handles English, Chinese, and mixed content

• Context Awareness: Understands code semantics and relationships

Search Types

1. Semantic Search (semantic): AI-powered understanding

2. Filename Search (filename): Fast filename matching

3. Extension Search (extension): Filter by file type

4. Hybrid Search (hybrid): Combines semantic + filename

Technical Stack

• Embedding Model: BAAI/bge-m3 (1024-dimensional vectors)

• Vector Database: ChromaDB

• Text Splitting: Intelligent semantic chunking

• Incremental Updates: Hash-based change detection

📁 Supported Formats

Auto-detected Text Files

• Programming languages: .py, .js, .ts, .java, .cpp, .c, .go, .rs, etc.

• Config files: .json, .yaml, .toml, .ini, .xml, .env

• Documentation: .md, .txt, .rst

• Web files: .html, .css, .scss

• Data files: .csv, .tsv

• Files without extensions (auto-detected)

Document Formats (Auto-converted to Markdown)

• Microsoft Office: .docx, .xlsx, .pptx

• OpenDocument: .odt, .ods, .odp

• PDF: .pdf (text extraction)

• Legacy formats: .doc, .xls (limited support)

🔒 Security Features

Access Control

• Directory Restriction: Access limited to SAFE_DIRECTORY and subdirectories

• Path Traversal Protection: Automatic prevention of ../ attacks

• Symlink Control: Configurable symbolic link access

• File Size Limits: Prevents reading oversized files

Validation

• Path Sanitization: Automatic path cleaning and validation

• Permission Checks: Verify read permissions before access

• Error Handling: Graceful failure with informative messages

🔗 Integration

Claude Desktop

Add to your Claude Desktop MCP configuration:

{
  "mcpServers": {
    "fs-mcp": {
      "command": "python",
      "args": ["main.py"],
      "cwd": "/path/to/fs-mcp",
      "env": {
        "SAFE_DIRECTORY": "/your/project/directory"
      }
    }
  }
}

Other MCP Clients

Connect to http://localhost:3002 using Server-Sent Events (SSE) protocol.

API Integration

The server exposes standard MCP endpoints that can be integrated with any MCP-compatible client.

🏗️ Project Structure

fs-mcp/
├── main.py                    # Main MCP server
├── src/                       # Core modules
│   ├── __init__.py           # Package initialization
│   ├── file_reader.py        # Core file reading logic
│   ├── security_validator.py # Security and validation
│   ├── text_detector.py      # Intelligent file detection
│   ├── config_manager.py     # Configuration management
│   ├── document_cache.py     # Document caching system
│   ├── file_converters.py    # Document format converters
│   ├── dir_tree.py          # Directory tree generation
│   ├── embedding_config.py   # AI embedding configuration
│   ├── codebase_indexer.py   # Vector indexing system
│   ├── codebase_search.py    # Search engine
│   ├── index_scheduler.py    # Index scheduling
│   └── progress_bar.py       # Progress display utilities
├── tests/                    # Test suite
├── cache/                    # Document cache (auto-created)
├── logs/                     # Log files (auto-created)
├── pyproject.toml           # Project configuration
├── .env.example             # Environment template
├── .gitignore              # Git ignore rules
└── README.md               # This file

💻 Development

Setting Up Development Environment

# Clone repository
git clone https://github.com/yourusername/fs-mcp.git
cd fs-mcp

# Install with development dependencies
uv sync --group dev

# OR with pip
pip install -e ".[dev]"

Running Tests

# Run all tests
pytest

# Run with coverage
pytest --cov=src

# Run specific test
pytest tests/test_file_reader.py

Code Quality

# Format code
black src/ tests/

# Lint code
flake8 src/ tests/

# Type checking
mypy src/

Debugging

Monitor logs in real-time:

tail -f logs/mcp_server_$(date +%Y%m%d).log

🤝 Contributing

We welcome contributions! Here's how to get started:

1. Fork and Clone

git clone https://github.com/yourusername/fs-mcp.git
cd fs-mcp

2. Create Feature Branch

git checkout -b feature/your-feature-name

3. Make Changes

• Follow the existing code style

• Add tests for new functionality

• Update documentation as needed

4. Test Your Changes

pytest
black src/ tests/
flake8 src/ tests/

5. Submit Pull Request

• Describe your changes clearly

• Reference any related issues

• Ensure all tests pass

Development Guidelines

• Code Style: Follow PEP 8, use Black for formatting

• Testing: Maintain test coverage above 80%

• Documentation: Update README and docstrings

• Commits: Use conventional commit messages

• Security: Follow security best practices

📋 Roadmap

• Enhanced PDF Processing: Better table and image extraction

• More Embedding Models: Support for local models

• Real-time Indexing: File system watchers

• Advanced Search: Regex, proximity, faceted search

• Performance Optimization: Async processing, caching improvements

• Web Interface: Optional web UI for management

• Plugin System: Custom file type handlers

• Enterprise Features: Authentication, rate limiting, monitoring

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

🙏 Acknowledgments

• FastMCP - MCP server framework

• LangChain - AI integration

• ChromaDB - Vector database

• BGE-M3 - Embedding model

📞 Support

• Issues: GitHub Issues

• Discussions: GitHub Discussions

• Documentation: Check the docs/ folder (when available)

Related MCP server: MCP Filesystem Server

中文

🚀 功能特点

• 🧠 智能文本检测: 无需依赖扩展名，自动识别文本文件

• 📄 多格式支持: 支持文本文件和文档格式（Word、Excel、PDF等）

• 🔒 安全验证: 只允许读取配置的安全目录中的文件

• 📏 按行读取: 支持指定行范围读取，便于处理大文件

• 🔄 文档转换: 自动将文档格式转换为Markdown并缓存

• 🔍 向量搜索: 基于AI嵌入的语义搜索

• ⚡ 高性能: 支持批量文件处理和智能缓存

• 🌐 多语言: 支持中英文内容处理

🚀 快速开始

1. 克隆和安装

git clone https://github.com/yourusername/fs-mcp.git
cd fs-mcp

# 推荐使用 uv
uv sync

# 或使用 pip
pip install -r requirements.txt

2. 环境配置

创建 .env 文件：

# 安全设置
SAFE_DIRECTORY=.                    # 目录访问限制（必需）
MAX_FILE_SIZE_MB=100                # 文件大小限制（MB）

# 编码设置
DEFAULT_ENCODING=utf-8

# AI嵌入配置（用于向量搜索）
OPENAI_EMBEDDINGS_API_KEY=your-api-key
OPENAI_EMBEDDINGS_BASE_URL=http://your-embedding-service/v1
EMBEDDING_MODEL_NAME=BAAI/bge-m3    # 或您偏好的模型
EMBEDDING_CHUNK_SIZE=1000

3. 启动服务器

python main.py

服务器将在 http://localhost:3002 启动并自动建立向量索引。

🛠️ MCP工具说明

详细的工具使用方法请参考英文部分的 MCP Tools 章节。

🔍 向量搜索功能

• 概念匹配：搜索"用户认证"能找到"登录验证"相关代码

• 同义词理解：搜索"database"能找到"数据库"相关内容

• 多语言支持：同时理解中英文代码和注释

• 上下文理解：理解代码的语义和上下文关系

📁 支持的文件格式

详细的格式支持请参考英文部分的 Supported Formats 章节。

🔒 安全特性

• 路径验证: 只允许访问配置的安全目录及其子目录

• 文件大小限制: 防止读取过大文件

• 路径遍历防护: 自动防止 ../ 等路径遍历攻击

• 符号链接控制: 可配置是否允许访问符号链接

🔗 集成方式

Claude Desktop集成

在 Claude Desktop 的 MCP 配置中添加：

{
  "mcpServers": {
    "fs-mcp": {
      "command": "python",
      "args": ["main.py"],
      "cwd": "/path/to/fs-mcp",
      "env": {
        "SAFE_DIRECTORY": "/your/project/directory"
      }
    }
  }
}

💻 开发

开发环境设置

# 克隆仓库
git clone https://github.com/yourusername/fs-mcp.git
cd fs-mcp

# 安装开发依赖
uv sync --group dev

运行测试

# 运行所有测试
pytest

# 运行覆盖率测试
pytest --cov=src

🤝 贡献

欢迎贡献代码！请参考英文部分的 Contributing 章节了解详细信息。

📄 许可证

本项目采用 MIT 许可证 - 详见 LICENSE 文件。

Made with ❤️ for the AI community

⬆ Back to top