📝 AI Sticky Notes - Python MCP Server

A lightweight, client-agnostic Model Context Protocol (MCP) server built in Python using FastMCP. This server exposes persistent note-taking capabilities to AI assistants (such as Claude Desktop, Google Antigravity, Cursor, and any MCP-compliant client).

---

🌟 Key Features

This server implements the three core primitives of the Model Context Protocol:

• 🛠️ Tools (Action Execution):
  • add_note(message: str) -> str: Appends a new sticky note to the local storage file.
  • read_notes() -> str: Reads and returns all saved notes formatted as a single string.
• 📦 Resources (Context Feeds):
  • notes://latest: Exposes the most recently saved note dynamically over a custom URI.
• 💬 Prompts (Reusable LLM Templates):
  • note_summary_prompt(): Generates a ready-to-use prompt directing the AI to summarize all current sticky notes.

---

📂 Project Structure

.
├── main.py                  # Primary FastMCP server implementation
├── main_final.py            # Clean reference implementation
├── notes.txt                # Persistent note storage (auto-generated)
├── pyproject.toml           # uv / Python package and dependency configuration
├── BITACORA_DE_ERRORES.md   # Troubleshooting log & root-cause analyses (Linux & PATH notes)
├── MCP_PYTHON_SDK.md        # FastMCP quick reference cheatsheet
└── README.md                # Project documentation (this file)

---

🚀 Getting Started

Prerequisites

• Python 3.12+
• uv (recommended for fast package and environment management)

1. Installation

Clone this repository and sync the dependencies:

git clone https://github.com/<your-username>/<your-repo-name>.git
cd <your-repo-name>
uv sync

---

🧪 Running & Testing

Option A: Interactive MCP Inspector (Recommended)

The MCP Inspector provides a visual web interface to test tools, inspect resources, and execute prompts interactively:

uv run mcp dev main.py

Open the URL displayed in the terminal (usually http://localhost:5173 or similar) to interact with your server.

---

Option B: Claude Desktop Integration

Add the server definition to your claude_desktop_config.json:

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
• Windows: %APPDATA%\Claude\claude_desktop_config.json
• Linux: ~/.config/Claude/claude_desktop_config.json

{
  "mcpServers": {
    "ai-sticky-notes": {
      "command": "uv",
      "args": [
        "--directory",
        "/absolute/path/to/your/project",
        "run",
        "main.py"
      ]
    }
  }
}

💡 Linux / GUI Environment Tip: If your desktop environment does not inherit your shell's interactive PATH (causing spawn uv ENOENT), specify the absolute path to your uv binary (e.g. /home/<user>/.local/bin/uv or /home/<user>/miniconda3/bin/uv). See BITACORA_DE_ERRORES.md for details.

---

Option C: Integration with Standard MCP Clients (Cursor, Antigravity, Custom CLI)

Configure standard stdio transport using:

• Command: uv
• Args: ["--directory", "<PROJECT_PATH>", "run", "main.py"]

---

📖 MCP API Reference

Tools

Name	Parameters	Return Type	Description
add_note	message: str	str	Adds a new note line to notes.txt
read_notes	(none)	str	Returns all recorded notes or "No notes yet"

Resources

URI Pattern	Name / Description	Return Content
notes://latest	Get Latest Note	Text content of the last recorded note

Prompts

Name	Arguments	Description
note_summary_prompt	(none)	Returns a prompt asking the AI model to summarize the current notes

---

🛠️ Development & Troubleshooting

For a detailed changelog of debugging steps, Linux platform workarounds, and common setup issues encountered during development, refer to:

• 📓 BITACORA_DE_ERRORES.md (Log in Spanish)
• 📘 MCP_PYTHON_SDK.md (SDK Cheatsheet)

---

📄 License & Credits

• Built with Model Context Protocol and FastMCP.
• Inspired by the MCP server development tutorial by Tech With Tim.