ai-distiller-mcp
Provides filesystem access to Claude via the MCP protocol, enabling local file operations through natural language.
README
My MCP for Filesystem Access for Claude
References
TL;DR
Only works in local Claude for now.
TODO: dxt package is too big to be added as an extension. Fix this
To test it locally:
uv pip compile pyproject.toml > requirements.txtuv run mcp-server --debug
To run it in Claude, add this to your claude_config.json file:
"ai_distiller": {
"command": "uv",
"args": [
"--directory", "/Users/fperez/dev/ai-distiller-mcp",
"run",
"mcp-server"
]
},
Dev
Requirements
pixiuv
Install dev env
pixi clean ; pixi install
Run local server for dev purposes
pixi run python -m server.main --debug
Testing it with Anthropic MCP Inspector
NOTE: Shutdown the local server if you started it
Open the inspector:
DANGEROUSLY_OMIT_AUTH=true npx @modelcontextprotocol/inspector
And then:
- go to the URL specified
- select stdio as connection method
- put
pixiinCommandinput box - put
run python server/main.py --debugin theArgumentsinput box - click connect
- go to tools and list them
NOTE For testing the server as extension is run when added to Claude try:
PYTHONPATH="/Users/fperez/Library/Application Support/Claude/Claude Extensions/local.dxt.francisco-perez-sorrosal.ai-distiler-mcp/server/lib" python /Users/fperez/Library/Application\ Support/Claude/Claude\ Extensions/local.dxt.francisco-perez-sorrosal.ai-distiler-mcp/server/main.py --debug
Packaging MCP server as DXT
TODO!!!
Init the dxt
TODO This does not implemented yet
Init dxt project with a manifest
npx @anthropic-ai/dxt init --yes
Note When creating the manifest, in the mpc_config section, put the full path to the python interpreter -> "command": "/Users/fperez/.pyenv/shims/python"
Bundle Python libs and Package Project
TODO This is not implemented yet
pixi install
pixi run bundle
pixi run pack
The output .dxt file is created on the dxt-package directory. Once the .dxt file is created you can drag and drop it to Claude (in the Settings/Extensions section.)
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。