MCP Local File Saver

A Python-based MCP server for Claude Desktop that lets you save, append, read, and list local text notes.

Features

• Save a note as a local .txt file
• Append new content to an existing note
• Append Claude’s response to a note
• Read a note by title
• List all saved notes
• Expose saved notes through an MCP resource
• Summarize note content with an MCP prompt

Built With

• Python
• MCP Python SDK
• Claude Desktop
• uv

Project Structure

mcp-local-file-saver/
├── main.py
├── .gitignore
├── pyproject.toml
├── uv.lock
└── notes/   # ignored from git

MCP Capabilities

Tools

• save_note(title, content)
• append_note(title, content)
• append_response_to_note(note_name, response_text)
• read_note(title)
• list_notes()

Resources

• notes://all

Prompts

• summarize_note(...)

Example Use Cases

You can ask Claude things like:

• Save this response to mcp-learning
• Append that answer to oracle-notes
• Read my fav players note
• List all my saved notes
• Summarize my interview prep note

Setup

1. Clone the repo

git clone <your-repo-url>
cd mcp-local-file-saver

2. Install dependencies

uv sync

3. Run locally with MCP Inspector

uv run mcp dev main.py

4. Run with Claude Desktop

Add the server to your Claude Desktop config using your local Python interpreter inside .venv.

Example:

{
  "mcpServers": {
    "Local File Saver": {
      "command": "/absolute/path/to/project/.venv/bin/python",
      "args": [
        "/absolute/path/to/project/main.py"
      ]
    }
  }
}

Notes

• Notes are stored locally in the notes/ folder
• The notes/ folder is ignored in git
• The server runs over stdio

Why I Built This

I built this project to get hands-on experience with MCP by creating a practical local productivity tool. It helped me understand how MCP tools, resources, and prompts work together in a real Claude Desktop integration.

Future Improvements

• Support Apple Pages
• Support Google Docs
• Search notes by keyword
• Read notes by exact filename
• Add tags or metadata to notes
• Add better summarization workflows

License

MIT