aiogram-mcp 0.2.0

MCP server middleware for aiogram Telegram bots — expose your bot to AI agents via the Model Context Protocol

aiogram-mcp

MCP server middleware for aiogram Telegram bots.

aiogram-mcp lets you expose an existing aiogram bot to MCP clients such as Claude Desktop without rewriting handlers, routers, or business logic.

Status

Beta — the core API is stable but may change before 1.0.

Installation

pip install aiogram-mcp

Requirements:

• Python 3.10+
• aiogram 3.20+

Quickstart

from aiogram import Bot, Dispatcher
from aiogram_mcp import AiogramMCP

bot = Bot(token="YOUR_BOT_TOKEN")
dp = Dispatcher()

# Register your normal handlers here.

mcp = AiogramMCP(bot=bot, dp=dp)
await mcp.run_alongside_bot(transport="stdio")

Available transports:

• stdio for Claude Desktop and local MCP clients
• sse for remote HTTP-based MCP connections

Built-in Tools

Messaging:

• send_message
• send_photo
• forward_message
• delete_message
• pin_message

Users:

• get_bot_info
• get_chat_member_info
• get_user_profile_photos

Chats:

• get_chat_info
• get_chat_members_count
• ban_user
• unban_user
• set_chat_title
• set_chat_description

Broadcast:

• broadcast when enable_broadcast=True

MCP Resources

Read-only data exposed to AI agents without tool calls:

URI	Description
telegram://bot/info	Bot metadata (username, capabilities)
telegram://config	Server configuration and allowed chat IDs
telegram://chats	Active chats the bot has seen (requires middleware)
telegram://chats/{chat_id}/history	Recent message history for a chat

Resources require MCPMiddleware to be attached for chat tracking and message history.

Safety Controls

mcp = AiogramMCP(
    bot=bot,
    dp=dp,
    name="my-bot",
    allowed_chat_ids=[123456789, -1001234567890],
    enable_broadcast=True,
    max_broadcast_recipients=500,
)

Use MCPMiddleware to track chats, users, and message history for MCP resources:

from aiogram_mcp import AiogramMCP, MCPMiddleware

tracker = MCPMiddleware(history_size=50)
dp.message.middleware(tracker)

mcp = AiogramMCP(bot=bot, dp=dp, middleware=tracker, enable_broadcast=True)

Development

pip install -e ".[dev]"
ruff check aiogram_mcp tests examples
mypy aiogram_mcp
pytest -v

Project layout:

• aiogram_mcp/ package source
• tests/ unit tests
• examples/ runnable usage examples
• .github/workflows/ci.yml GitHub Actions pipeline

Examples

• basic_bot.py
• incident_alert_bot.py

License

MIT. See LICENSE.