SitePilot (MCP-Powered Website Operations)

SitePilot is a Telegram-controlled, AI-assisted static website maintenance system. You send a request from Telegram, and the project uses an MCP-style operation loop to read/edit files in site/, then deploy changes through Git.

What this project does

• Accepts website change requests from Telegram (/pilot ...)
• Uses a local LLM (via Ollama) to generate structured file operations
• Executes file operations inside the site/ folder
• Auto-commits and pushes updates for deployment (for example, GitHub Pages)
• Optionally sends Telegram deployment notifications

MCP role in this project

The file mcp_server.py is the core operation engine.

It works like an MCP-style controller by:

• receiving a user task
• exposing available project files to the model
• asking the model to return explicit commands only
• parsing and executing those commands deterministically

Supported operation commands

The model is instructed to return only these operations:

• READ: <filename> - read a file from site/
• FILE: <filename> + content block - create/update a file in site/
• DELETE: <filename> - delete a file from site/

This command-contract is the project's MCP-like interface between AI planning and actual file execution.

Architecture

• telegram_bot.py Telegram bot entrypoint and command handlers (/start, /pilot, /status).
• mcp_server.py MCP-style operation server: model prompt, command parsing, filesystem operations, Git deploy, and notifications.
• site/ The actual website files that the AI edits.
• requirements.txt Python dependencies.

End-to-end workflow

1. User sends /pilot <change request> in Telegram.
2. Bot launches:

   python mcp_server.py "<change request>"

3. mcp_server.py gathers current site/ structure.
4. Ollama model returns operation commands (READ, FILE, DELETE).
5. Server executes commands against site/.
6. Server runs Git deploy steps (add, commit, push).
7. Optional Telegram notification is sent.

Workflow Diagram

Prerequisites

• Python 3.12+
• Ollama installed locally
• Ollama model available locally: qwen2.5-coder:7b
• Telegram bot token from @BotFather
• Git remote configured (if you want automatic push/deploy)

Installation

git clone <your-repo-url>
cd MCP
pip install -r requirements.txt

Environment variables

Create a .env file in the project root:

TELEGRAM_TOKEN=your_bot_token_here
TELEGRAM_NOTIFY_IDS=your_chat_id

• TELEGRAM_TOKEN: required for bot + notifications
• TELEGRAM_NOTIFY_IDS: optional comma-separated chat IDs to notify after deploy

Run the bot

python telegram_bot.py

Then in Telegram:

• /start - usage help
• /status - health check
• /pilot Add a dark mode toggle to the homepage

Direct MCP operation run (without Telegram)

You can run the MCP server directly:

python mcp_server.py "Update hero section text and CTA button"

Demo

Video: https://youtu.be/wWsGnYfuECk