mcp-file-roots-demo
Enables interactive chat with AI models via Anthropic API, with controlled file system access to specified directories and video conversion capabilities.
README
MCP Chat with File System Access
MCP Chat is a command-line interface application that enables interactive chat capabilities with AI models through the Anthropic API. The application supports file system operations with controlled access to specified directories, video conversion capabilities, and extensible tool integrations via the MCP (Model Control Protocol) architecture.
Prerequisites
- Python 3.10–3.13 (Python 3.14+ is not supported — dependencies lack pre-built wheels for it)
- Anthropic API Key
- FFmpeg (for video conversion features)
Setup
You must have FFmpeg already installed to convert a video file. To install FFmpeg on MacOS run:
brew install ffmpeg
Step 1: Configure the environment variables
- Copy the
.env.examplefile to create a new.envfile:
cp .env.example .env
- Edit the
.envfile and set your environment variables:
CLAUDE_MODEL="claude-sonnet-4-6" # Or your preferred Claude model
ANTHROPIC_API_KEY="" # Enter your Anthropic API secret key
Step 2: Install dependencies
Setup with uv
uv is a fast Python package installer and resolver.
- Install uv, if not already installed:
pip install uv
- Pin Python to a supported version (3.13 recommended) and install dependencies:
uv python pin 3.13
uv sync
- Run the project
When running the project, you must specify one or more root directories that the MCP server will have access to. Only files and directories within these roots can be accessed by the server.
uv run main.py <root1> [root2] [root3] ...
Examples:
# Single directory
uv run main.py /path/to/videos
# Multiple directories
uv run main.py /home/user/videos /mnt/storage/media ~/Documents
# Current directory
uv run main.py .
Features
File System Access
The server can only access files and directories within the specified root paths. This provides security by limiting file system access to approved locations.
Available Tools
- list_roots: List all accessible root directories
- read_dir: Read contents of a directory (must be within a root)
- convert_video: Convert MP4 videos to other formats (avi, mov, webm, mkv, gif)
Video Conversion
The video conversion tool uses FFmpeg to convert MP4 files to various formats:
- Standard video formats: AVI, MOV, WebM, MKV
- GIF conversion with optimized settings
- Medium quality preset for balanced file size and quality
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。