Telegram MCP — Watcher, Parser & Analysis

Full-featured Telegram MTProto MCP server. Connects to Telegram as a userbot and exposes message reading, sending, and parsing as MCP tools.

Features

• Userbot watcher — monitors all messages, triggers AI agent on keyword
• Chat/DM parser — paginated parsing of any chat or DM via parse_chat.py
• Context builders — stats, top words, LLM summaries, people profiles
• Criminal analysis — detects CSAM, fraud, NCA violations, hate speech
• Personality profiling — psychological portrait from message history

Setup

cp .env.example .env
# Fill in API_ID, API_HASH, SESSION_NAME, OWNER_ID from https://my.telegram.org/apps
pip install -r requirements.txt
python3 session_string_generator.py  # authorize session once

Run watcher as systemd service

sudo cp tg-watcher.service /etc/systemd/system/
sudo systemctl enable --now tg-watcher

MCP Tools (via mcporter)

• send_message — send message to chat
• reply_to_message — reply to specific message
• iter_all_messages — paginated message fetch
• iter_messages_by_user — messages from specific user
• and 90+ more Telegram API methods

Important notes

See CLAUDE.md for full list of gotchas. Key ones:

• Use \n not <br> in HTML messages
• iter_all_messages for DMs, not list_messages
• Semaphore required for parallel agent calls

Source

GitHub: https://github.com/SkrudjReal/telegram-mcp